Liu et al.

Soft Sci 2022;2:22

DOI: 10.20517/ss.2022.23
Soft Science

Review Article Open Access

Recent progress in flexible piezoelectric devices

toward human-machine interactions
Jiaqing Liu, Guo Tian, Weiqing Yang, Weili Deng*

Key Laboratory of Advanced Technologies of Materials (Ministry of Education), School of Materials Science and Engineering,
Southwest Jiaotong University, Chengdu 610031, Sichuan, China.

*Correspondence to: Dr. Weili Deng, Key Laboratory of Advanced Technologies of Materials (Ministry of Education), School of
Materials Science and Engineering, Southwest Jiaotong University, No. 111, North 1st Section of Second Ring Road, Jinniu District,
Chengdu 610031, Sichuan, China. E-mail: [email protected]

How to cite this article: Liu J, Tian G, Yang W, Deng W. Recent progress in flexible piezoelectric devices toward human-machine
interactions. Soft Sci 2022;2:22. https://dx.doi.org/10.20517/ss.2022.23

Received: 28 Sep 2022 First Decision: 31 Oct 2022 Revised: 21 Nov 2022 Accepted: 8 Dec 2022 Published: 19 Dec 2022

Academic Editors: Zhifeng Ren, Chuan Fei Guo Copy Editor: Fangling Lan Production Editor: Fangling Lan

Abstract
Human-machine interactions are becoming increasingly required for intelligent sensing and effective manipulation.
Recent developments in flexible piezoelectric sensors with short response time and high force-electric
interconversion efficiency present a tendency toward facilitating diverse human-machine interactive applications.
Here, we review the development of flexible piezoelectric human-machine interactions in the context of robotic
control, the Internet of Things, sports coaching and acoustic therapeutics. The synthesis of unique materials, the
distinct design of device structures, the typical applications of piezoelectric human-machine interactions and the
integration of cutting-edge technologies are elaborated in detail based on recent research. Finally, we highlight the
current challenges and directions for the development of piezoelectric human-machine interactions for more
advanced application scenarios.

Keywords: Piezoelectric devices, human-machine interactions, wearable devices, flexible electronics

INTRODUCTION
With the rapid development of sensing and information technologies, smart flexible electronics have
become a popular research topic, especially concerning cutting-edge technologies like artificial intelligence
(AI) and human-machine interactions (HMIs), which greatly facilitate and enrich people’s lives[1-4]. Effective

© The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0
International License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, sharing,
adaptation, distribution and reproduction in any medium or format, for any purpose, even commercially, as
long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and
indicate if changes were made.

www.softscijournal.com
Page 2 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

overall design and continuous improvement of user experiences are key to the development of HMI
systems[5-7]. Nevertheless, conventional electronic devices are often rigid and unwieldy, restricting their
wearability and conformability with human skin, which severely affects the sensing accuracy and user
experiences. On this basis, many researchers globally have investigated flexible electronic devices with more
eminent stretchability and flexibility for applications such as health monitoring[8-14] and gesture
interactions[15-17]. These flexible devices can be fixed to any shape of object[18,19], such as skin[20-23], gloves[24,25]
and rackets[26], to continually monitor human behavior in real time. HMIs serve as a bridge for information
exchange between users and machines and also represent an indispensable link for next-generation
applications, such as virtual reality (VR)[27-29]. Generally, a typical HMI system can be categorized into
several parts, including signal perception, information transmission, signal processing and feedback
execution. Among these parts, signal perception is the most important part as it detects external stimuli,
especially tactile signals, and builds the basis of the whole interaction process.

To date, several principles have often been utilized for sensing in HMI systems, including the
piezoelectric[20,26,30,31], piezoresistive[32,33], capacitive[34,35] and triboelectric effects[19,25,36]. Among them, human-
machine interfacial sensors based on the piezoelectric effect, which can generate an electric displacement in
response to an applied mechanical stimulus, have been widely used for tactile sensing, due to their
outstanding performance with high sensitivity, fast response time and self-powered operation[37].
Furthermore, due to the inverse piezoelectric effect, piezoelectric materials can be deformed under an
external electric field, which facilitates the preparation of an all-in-one device with both sensing and
actuation functions. Moreover, incorporated with various novel technologies like augmented reality (AR),
virtual reality and machine learning, piezoelectric HMIs (P-HMIs) are expected to improve people’s daily
lives to a more intelligent, convenient and interconnected level[38-41].

This review covers the recent progress made regarding smart flexible piezoelectric devices in the field of
HMIs. Research concerning flexible piezoelectric electronics is comprehensively summarized and
highlighted along with real-life P-HMIs, including robotic control, the Internet of Things (IoT), sports
coaching, acoustic therapeutics and machine learning-enhanced P-HMIs [Figure 1][27,42-46]. Finally, the
current challenges and prospects of P-HMIs are provided from our perspective, ranging from materials
science and structural design to diverse potential applications in the near future.

HMIS OF FLEXIBLE PIEZOELECTRIC DEVICES

P-HMIs for robotic control
Recently, with increasing demands for immersion in virtual interactions, the term “metaverse” has become a
heated topic in both academia and industry. With the help of smart flexible devices, combined with VR and
AR technologies, people can experience immersive interactions in the virtual world, such as playing video
games, taking online courses and attending distant conferences[47,48]. In contrast, some extremely harsh
environments, such as high temperatures and pressures and confined spaces, make it difficult to approach
for manipulation. Therefore, it is of great significance to control robotic systems remotely and interact with
robotic systems actively to settle real-life challenges.

The interaction between robots and controllers is a rather complex process that includes highly sensitive
sensing, signal and data processing and the execution of control[43,49]. In particular, this process places high
demands on response speed and execution accuracy. Thus, a growing number of researchers are devoted to
recognizing and mimicking human motions and enabling the interaction with robots by improving the
performance of flexible piezoelectric devices. For example, Deng et al. designed a flexible self-powered
piezoelectric sensor based on unique cowpea-structured PVDF/ZnO nanofibers, which demonstrated
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 3 of 19

Figure 1. Overview of P-HMIs, including robotic control, the Internet of Things, sports coaching, acoustic therapeutics and machine
learning. Reproduced with permission from ref.[42]. Copyright 2015, Royal Society of Chemistry. Reproduced with permission
from ref.[43]. Copyright 2020, Elsevier. Reproduced with permission from ref.[27]. Copyright 2022, Elsevier. Reproduced with permission
from ref.[44]. Copyright 2021, IEEE. Reproduced with permission from ref.[45]. Copyright 2021, MDPI. Reproduced with permission
from ref.[46]. Copyright 2019, Wiley.

excellent mechanical and electrical properties for gesture interaction with a robotic hand[20]. When the
sensor was fixed on a finger knuckle, its bending movements could be converted into electrical signals by
the sensor to remotely perform control over the robotic hand via a controlling circuit. In this work, the
robotic hand demonstrated its ability to synchronously imitate the same gesture as the human hand, as
illustrated in Figure 2A. This self-powered sensor possesses a unique material structure and functioning
mechanism, thereby exhibiting promising potential for intelligent HMIs, like gesture recognition and
imitation.

Obviously, the effective design of functional materials and device structures enabling the regulation of the
overall performance and functionality of piezoelectric devices has played an indispensable role in the
development of HMIs. A common strategy is to use thin piezoelectric ceramics with excellent electrical
performance to make them more flexible or to dope piezoelectric polymers with good flexibility to improve
their electrical output. Recently, many two-dimensional piezoelectric materials have also gained attention
due to their lightweight and flexibility[50]. For example, Yang et al. reported a piezoelectric device based on
tin disulfide (SnS2) nanosheets, which showed excellent energy harvesting and active sensing capabilities[51].
As illustrated in Figure 2B, the device had a conducting channel between 2-5 nm-thick SnS2 nanosheets,
electrically connected by multiple Cr/Au electrodes. The SnS2 nanosheets are hexagonal crystals with an
Page 4 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

Figure 2. P-HMIs for robotic control. (A) Schematic diagram of HMI between a human hand and a remotely controlled robotic hand.
Reproduced with permission from ref.[20]. Copyright 2019, Elsevier. (B) Schematic illustration of PENG device with a channel of SnS2
electrically connected by Cr/Au electrodes on a polyimide substrate, where blue and green spheres represent Sn and S atoms,
respectively. (C) TEM image with single few-layered SnS2 nanosheets shown in the inset. Scale bar of 1 μm. Reproduced with
permission from ref.[51]. Copyright 2020, Elsevier. (D) FEA strain distributions of heterostructure device in stretched mode. (E) Output
voltage of sensor in stretched mode as a function of time and an image of sensor in stretched mode. Reproduced with permission
from ref.[54]. Copyright 2014, Wiley. (F) Image of using VR to grab an object in virtual space by detecting finger motions in the real
world. Reproduced with permission from ref.[40]. Copyright 2019, Springer Nature. (G) Schematic of human-machine interactive sensor
array system. (H) Output voltage of sensor as a function of time when using a robotic hand with a feedback module to grasp an object.
Reproduced with permission from ref.[55]. Copyright 2021, Wiley.

interplanar spacing of d(100) ¼ 0.32 nm[52,53], as the transmission electron microscopy (TEM) image shows
in Figure 2C. In this work, the sensor was successfully demonstrated to achieve the real-time gesture control
of a robotic hand by sign language, including a handshake between the robot and the human. The excellent
self-powering and dynamic sensing capabilities of the device show rich potential in multidisciplinary fields
for future applications, including energy harvesting, intelligent robots and HMIs.

In addition, Lim et al. fabricated a transparent and stretchable HMI system with wearable mechanical
sensors and stimulators, allowing the feedback to control the movements of a robot[54]. Due to the
prominent electrical and mechanical properties of the utilized conductive/piezoelectric graphene
heterostructures, the designed piezoelectric sensors exhibited great flexibility and high detection sensitivity,
thereby enabling the accurate monitoring of human motions. The strain distribution of the motion sensor
stimulated by FEA is displayed in Figure 2D, presenting the strain magnitudes and polarities in the
heterostructure of this device. When the wrist was bent, the stretch mode would be triggered and a unique
corresponding command signal would be generated to control the robotic hand to move [Figure 2E]. In
addition, more complex motion modes, such as radial deviation, ulnar deviation, pronation and supination,
could be detected by assembling several sensors into a sensor array. The successful development of the
above-mentioned smart devices allows for more sophisticated and intelligent HMIs.
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 5 of 19

With the advent of the metaverse, increasingly more people want to interact in a virtual world since it is not
limited by time and space. Sensors and actuators are the necessary bridge between virtual space and the real
world. Song et al. designed a smart glove consisting of PVDF-based piezoelectric sensors and flexible
pneumatic actuators for providing tactile feedback, as shown in Figure 2F[40]. With the gloves on, finger
motions could be monitored by piezoelectric sensors and the resulting piezoelectric outputs are transmitted
to control the virtual hand. When the virtual hand touches a virtual object, the actuators are activated and
give tactile feedback to the finger, thus enabling interaction with the virtual world. In this work, researchers
tested this device in a virtual chess scenario, where a tester successfully picked up the pieces in the virtual
world.

Furthermore, an active sensor array allows for more complicated and advanced applications of HMI
systems compared to single-sensor systems. For instance, Lin et al. demonstrated a multilayer skin-inspired
piezoelectric tactile sensor array with crosstalk-free row+column electrodes[55]. Integrated with a signal
processor and a designated computing algorithm, the e-skin-based sensor array could detect and distinguish
the magnitudes, positions and motion modes of external stimuli in real time. As schematically illustrated in
Figure 2G, this HMI system resembled the biological response process of human beings. Initially, the
mechanical stimuli would be converted to electrical signals by a PVDF-based sensing array due to the
piezoelectric effect. The electrical signals were then transmitted to the peripheral devices, such as
smartphones, computers and cloud computing platforms. After processing and analysis by specific
algorithms, parameters, such as the positions, magnitudes and modes of the stimuli, could be accurately
identified in real time. In this work, the authors achieved an interesting application by applying the
piezoelectric sensor array with a robotic hand in an industrial scenario. By adjusting the default threshold
value (4 V) of the sensor feedback, the robotic hand can correctly control the gripping force during grasping
without exceeding the set value, thus protecting the grasped object from being damaged, as displayed in
Figure 2H.

Due to the rapid response time and high sensing accuracy of piezoelectric sensors, P-HMIs have proven to
be indispensable in real-time robotic control and interactions[56]. Incorporated with 5G technologies,
information between humans and robots can be transmitted instantly, enabling many applications that
previously seemed impossible. For example, doctors will be able to control medical robotics to conduct
remote surgeries regardless of location, making treatment more efficient and reducing healthcare
imbalances around the world[57]. However, these precise applications raise serious challenges for the overall
design of the system to ensure rapid response time and sensing accuracy. The use of P-HMIs not only
makes the interaction with the robot more convenient and efficient but also provides the possibility of
realizing the HMI under harsh conditions, thereby reducing the risks that people may suffer.
P-HMIs for Internet of Things
Benefitting from improvements in advanced sensing and information technologies, piezoelectric flexible
sensors for the IoT have been demonstrated in numerous HMIs, such as smart homes[58,59], smart traffic[60]
and intelligent healthcare[61,62]. For instance, Song et al. reported a piezoelectric device that could be installed
as a pavement block on a sidewalk for energy harvesting[63]. This IoT system could be utilized to power
walkway signal indicators and enable the collection and processing of environmental data (such as humidity
and temperature), as illustrated in Figure 3A. When a person weighing 60 kg stepped on the system ten
times, a capacitor (1000 μF) could be charged to 55.6 mJ. Consequently, the harvester successfully turned on
a delineator on an actual sidewalk after 4.8 s and powered the IoT system without an external energy supply.
This study proves the great durability and energy harvesting capability of the designed piezoelectric device,
which shows promising prospects in human motion monitoring and traffic signal control for HMIs.
Page 6 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

Figure 3. P-HMIs for IoT. (A) Schematic diagram of an IoT HMI system controlled by a pavement block piezoelectric energy harvester.
Reproduced with permission from ref.[63]. Copyright 2019, Elsevier. (B) Conceptual design and possible applications of a smart pen
based on a piezoelectric energy harvester. Reproduced with permission from ref.[64]. Copyright 2021, Elsevier. (C) Photograph of flexible
PEH-based sensor arrays for HMI. (D) Working mechanism of flexible tactile sensor-based smart door lock. Reproduced with
permission from ref.[65]. Copyright 2022, Elsevier. (E) Schematic overview of a smart keyboard application using flexible piezoelectric
sensors. Reproduced with permission from ref.[27]. Copyright 2022, Elsevier. (F) Dynamic voltage output of three adult subjects when
typing the word “DIY” on the keyboard attached with PMNP textiles. Reproduced with permission from ref.[66]. Copyright 2021, Elsevier.
(G) Photograph of a wireless switch system to achieve the real-time smart home control of a table lamp, an air conditioner and an air
purifier. Reproduced with permission from ref.[67]. Copyright 2018, Elsevier.

In addition, many researchers are also interested in embedding smart P-HMI devices into daily items. For
example, Kim et al. designed a smart pen-based piezoelectric energy harvester, which could generate energy
in both the d31 and d33 modes during writing[64]. Two piezoelectric devices, a cantilever and an impact-type
device, were installed in the pen. The designed device showed excellent energy-harvesting abilities and was
able to generate sufficient power to operate a wireless sensor. It was also capable of detecting the users’
writing movements. Based on the above functions, this pen-based IoT device may be applied to many HMIs
in the future, such as information security, wireless communication and the detection of various
environmental and biological signals, as shown in Figure 3B.

Recently, Lv et al. successfully demonstrated a piezoelectric IoT device for smart door locks[65]. They first
fabricated an all-inorganic transparent Sm-doped Pb(Mg1/3Nb2/3)O3-PbTiO3 (Sm:PMN-PT) film on a mica
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 7 of 19

substrate via a simple one-step sol-gel process. The synthesized soft film showed an ultrahigh piezoelectric
coefficient d33 of 380 pm V-1 and superior flexibility. Furthermore, a 3 × 2 piezoelectric sensor array was
fabricated to detect the spatial distribution of exerted forces, as shown in Figure 3C. When the passwords
were entered on the sensing array, the corresponding output voltage would be detected by a multichannel
signal collector and then compared with the preset password in the system. If the input numbers matched
the password in the memory, the door would be unlocked and vice versa, as illustrated in Figure 3D. Such
flexible PMN-PT films are proving to possess significant potential for the authentication of smart unlocks.

Similar to locks, the keyboard is another common and popular device, eligible for information exchange
between humans and machines. With the escalation of computer-related security issues, criminals can easily
bypass security authentication systems by hacking passwords, causing serious problems for the privacy of
personal information. On this basis, Zhou et al. demonstrated a smart keyboard-based personalized
recognition system, as depicted in Figure 3E[27]. In this study, several volunteers were asked to type the same
words on the smart keyboard, with piezoelectric sensors attached to its keys. From the results of the
piezoelectric output, the authors found that the amplitudes, intervals and waveforms of the outputs varied
with volunteers, showing individual personal features. Therefore, it is possible to identify the user through
analyzing the signal patterns collected by the smart keyboard. Likewise, Su et al. conducted a thorough
study regarding a smart keyboard based on barium titanate-doped polyvinylidene fluoride (BTO/PVDF)
piezoelectric nanocomposite fibers[66]. In their study, polydopamine-modified nonwoven piezoelectric
(PMNP) textiles were constructed based on the BTO/PVDF fibers, whose unique structure endowed the
textiles with prominent piezoelectric output and mechanical stiffness. The PMNP textiles were then
attached on a keyboard to detect the patterns of finger tapping. As shown in Figure 3F, three sets of voltage
output were measured when each student entered the same word “DIY” on the keyboard attached with the
PMNP textiles. As can be seen from the plot, there are significant differences between the three collected
waveforms. Several biometric features could be obtained by analyzing the data, such as typing forces, finger
sizes and typing patterns. All the additional information from P-HMIs can be used as a valid identifier for
the personalized interactive authentication system, proving its advantages over traditional devices.

Furthermore, Kim et al. investigated an excellent piezoelectric material (optimized PZN0.25C ceramic) for
smart-home sensors that could be attached to floor tiles[67]. When stepping on the floor, a peak output
voltage of 42 V and a current of 11 µA were generated by the PZN0.25C ceramic. This energy could power a
wireless sensor unit to control the on/off of home appliances, such as stand lamps, air conditioners and air
purifiers in real time, as schematically shown in Figure 3G. Furthermore, such piezoelectric devices can
easily be placed on floor tiles at a large scale, which provides more possibilities for HMIs.

To conclude, in the era of the IoT, flexible piezoelectric sensors can be attached to numerous daily
necessities to realize HMIs and make life more intelligent in a vast amount of scenarios[68,69]. At present, we
have already experienced the convenience of P-HMIs in smart homes, including voice-controlled or tactile-
sensing devices, such as smart curtains, voice assistants and automatic lights. With the further advancement
of IoT technologies, including sensing, signal transmission and information feedback, perhaps newly
developed smart IoTs could be controlled by facial expressions, muscle gestures or even emotions, which
will make our daily lives more convenient and efficient.
P-HMIs for sports coaching
As the era of AI approaches, advanced technologies have been integrated into many aspects of our lives[70].
Various wearable sensing and interactive information systems are widely used for physical fitness
monitoring and training instruction in sports. Traditionally, the training of athletes is judged and recorded
Page 8 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

primarily through visual observations from coaches or camera-type equipment, which often lacks
immediate or precise feedback. With the aid of P-HMIs, sports training can be more scientific and efficient
due to the real-time monitoring and sports data analysis[71-73]. Recently, Tian et al. designed a distinct self-
powered piezoelectric sensor with a rich lamellar crystal baklava-structured PZT/PVDF for guiding table
tennis[74], as shown in Figure 4A. Based on the outstanding performance of the designed sensors (sensitivity
of 6.38 mV/N and response time of 21 ms), the fabricated table tennis racket, attached with a 6 × 6 sensor
array, could accurately detect the hit location and the magnitude of the contact force. On this basis, coaches
would be able to provide customized guidance for the training of athletes from the data collected in real
time. The design of this smart racket offers promising prospects for the application of P-HMIs in individual
sports guidance and intelligent sports equipment.

In addition to racket games, like tennis and table tennis, P-HMIs can also be used in other sports
concerning continuous motion monitoring. Lu et al. designed a flexible self-powered piezoelectric sensor
for speed skating motion monitoring[45]. This sensor mainly consists of polarized PVDF, silver electrodes
and a flexible polyester substrate, which could be attached to the hip joint of skaters and the blades of
skating shoes [Figure 4B]. When skating, the device could generate corresponding piezoelectric voltages
with different amplitudes and durations in response to body movements, such as joint bending angles,
motion frequencies and skate blade vibrations, as shown in Figure 4C. These data are then stored and
analyzed to provide individual guidance for skaters.

Moreover, P-HMIs have also proven their potential in coaching other kinds of sports, such as boxing and
swimming. As shown in Figure 4D, Yao et al. embedded a 3D-printed flexible piezoelectric lattice into a
boxing glove to detect the magnitude and spatial distribution of punch force[75]. The obtained piezoelectric
signals could be instantly transmitted to cell phones via a WiFi module and paired with a graphical display.
When a person hit the wall with this glove, whether it was a direct hit or a right hook, the corresponding
spatial distribution of the force could be displayed on the cell phone, as shown in Figure 4E. As a result,
users can obtain feedback regarding their boxing performance simply from their mobile phones and use the
information to hone their boxing skills. Interestingly, Mao et al. demonstrated a flexible self-powered
biosensor based on ZnO nanowire arrays and a flexible polyethylene terephthalate (PET) substrate for the
real-time monitoring of swimming[76]. Due to the piezoelectric effect of polar ZnO nanowires, the sensor
could convert the biomechanical energy of swimming into electric energy without the need for an external
power supply in the water. As illustrated in Figure 4F, the sensor could be easily fixed on the skin to
monitor motions like joint moving angle and frequency during swimming. The sensor mainly consisted of
the planar electrode and flat ZnO nanowires, as shown in Figure 4G. The output voltage waveform of the
sensor allowed us to obtain and evaluate different swimming patterns [Figure 4H]. Likewise, the collected
signals could be transmitted to a specific terminal for further analysis to help people improve their
swimming skills.

It is generally believed that wearable P-HMIs will be applied to many forms and carriers, including but not
limited to gloves, shoes and e-skin. With the fast advancement of flexible electronics and AI, there will be
more state-of-the-art signal processing and machine learning algorithms to analyze sports-related data.
These advanced algorithms are expected to be integrated with expertise in sports science and medicine to
make sports training more scientific and personalized.
P-HMIs for acoustic therapeutics
With the surging need for personal medical treatment and rehabilitation, various therapeutic devices have
been developed. In particular, for implantable devices, power supplies, information exchange and feedback
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 9 of 19

Figure 4. P-HMIs for sports coaching. (A) Schematic diagram of smart racket and sensor structure for interactive table tennis coaching.
Reproduced with permission from ref.[74]. Copyright 2019, Elsevier. (B) Photograph of piezoelectric sensor attached to an ice skate
blade. (C) Measurement system for mimicking the motion of a knee joint. Reproduced with permission from ref.[45]. Copyright 2021,
MDPI. (D) Schematic diagram of P-HMI system based on a boxing glove. (E) Spatial distribution of force magnitudes obtained from
electrodes attached to boxing glove. Reproduced with permission from ref.[75]. Copyright 2019, Wiley. (F) Schematic diagram of
monitoring swimming athlete’s elbow joint angle and heart rate. (G) SEM image of a single ZnO nanowire crossing a pair of electrodes.
Scale bar of 10 μm. (H) Waveform of piezoelectric voltage output of butterfly stroke. Reproduced with permission from ref.[76].
Copyright 2021, MDPI.

control are crucial challenges that need to be addressed. In this process, acoustic waves not only transmit
energy wirelessly for therapeutic devices implanted in human bodies but also act as carriers to exchange
information for HMIs, which is an important medium recognized by researchers[77-81]. Piezoelectric devices
play an extraordinary role in the field of acoustic therapeutics due to their efficient bidirectional
electromechanical coupling capability. For example, Zhang et al. reported a piezoelectric ultrasound energy-
harvesting (PUEH) device that could be implanted underneath the skin for deep brain stimulation by
converting ultrasound energy to electrical energy [Figure 5A][82]. As schematically illustrated in Figure 5B,
the designed harvester was a multilayer structured device with a Sm-doped Pb(Mg1/3Nb2/3)O3-PbTiO3
(Sm-PMN-PT) single crystal as the functional layer. The Sm-PMN-PT crystal was laid at a thickness of 380
μm and then divided into small pieces with dimensions of 1 mm × 1 mm. These pieces were then integrated
with several flexible electrodes to form a 6 × 6 array and encapsulated in PDMS for excellent
biocompatibility and flexibility. In the experiment, the device implanted in a rat brain could generate an
output power of 280 μW under ultrasound stimulation, which could instantly stimulate the periaqueductal
gray brain area for analgesia.

In addition to analgesia, a plethora of people are also suffering from disabilities such as blindness and the
significance of electronic ocular prostheses to restore vision through neural electrical stimulations has
become apparent. Jiang et al. reported a flexible ultrasound-induced retinal stimulating piezo-array
(F-URSP) for evoking visual perception in blind individuals[83]. The prepared F-URSP, as an artificial ocular
prosthesis, had a two-dimensional piezo-array with 32-pixel stimulating electrodes to perceive images. Each
piezo-element could be respectively activated by ultrasound, which was individually transduced by
Page 10 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

Figure 5. P-HMIs for acoustic therapeutics. (A) Schematic diagram of Sm-PUEH device for deep brain stimulation and analgesia.
(B) Structural diagram of Sm-PUEH. Reproduced with permission from ref.[82]. Copyright 2022, American Association for the
Advancement of Science. (C) Schematic diagram showing working model of F-URSP, in which a focused ultrasound probe was used for
selective excitation of the piezo-array and the output voltage magnitude of each element in the array under the excitation of a single
focused probe. Reproduced with permission from ref.[83]. Copyright 2022, Springer Nature. (D) Diagram of StimDust system implanted
in a rat to stimulate its sciatic nerve. Scale bar of 1 mm. (E) Block diagram overview of neuro-stimulating system. Reproduced with
permission from ref.[84]. Copyright 2020, Springer Nature Limited. (F) Highly sensitive frequency response of PMAS for full-cover
phonetic spectrum using an ultrathin polymer, a piezoelectric membrane and a multichannel electrode. (G) Multiresonant
displacements of ultrathin PMAS membrane measured by an LDV under frequency sweep from 100 Hz to 4 kHz. (H) Smartphone
biometric authentication using machine learning-enhanced acoustic module integrated with mini PMAS, machine learning processor
and wireless transmitter. Reproduced with permission from ref.[85]. Copyright 2021, American Association for the Advancement of
Science.

programmable beamlines, resulting in particular piezoelectric patterns. As illustrated in Figure 5C, the
piezo-element in perpendicular contact with the ultrasound beam exhibited a high output voltage above the
threshold with a high signal-to-noise ratio of ~32.6 dB, while the other piezo-elements displayed discernable
diminution in the magnitude of output voltage because of the lower intensities of the received ultrasound.
As a consequence, the ultrasound-induced pattern reconstruction was successfully demonstrated in murine
retinal tissue. Such modes of operation, which link ultrasound acoustic induction and retinal electrical
stimulation via piezoelectric devices, exhibit great potential in the field of restoring vision to the blind
through visual reconstruction.

Recently, Piech et al. demonstrated a wireless ultrasonically powered implantable neural stimulator, which
could be used for interactive communication between human beings and machines[84]. As illustrated in
Figure 5D, the designed system consisted of a piezoceramic transducer, an energy-storage capacitor and an
integrated circuit, which directly attached to the sciatic nerve of a rat to fulfill electrical stimulation. A single
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 11 of 19

ultrasonic link provided power to both the mote and bidirectional communication, achieving the
interaction between the mote and the external transceiver, as shown by the functional diagram in Figure 5E.
The uplink data could be demodulated to indicate whether a stimulus had occurred and the downlink data
could send programmed commands to control the generation of the stimulation pulses through decoding.
When implanted into the sciatic nerve of several rat models, the device engaged repeatable stimulation
according to a series of physiological responses, successfully achieving the closed-loop nerve stimulation
therapy.

In addition, flexible piezoelectric acoustic sensors have attracted substantial attention as essential
components for human-machine interactive applications, such as the voice user interface. For example,
Wang et al. designed a highly sensitive piezoelectric mobile acoustic sensor (PMAS) for biomimetic
frequency band control[85]. Figure 5F illustrates the device structure of this PMAS, which consisted of an
ultrathin PET substrate, a piezoelectric membrane and several multichannel interdigital electrodes. The
internal residual stress of the PZT membrane was customized to reinforce dipole alignment among the
interdigital electrodes, which greatly increased the sensitivity of the device. As a result, the multiresonant
band of PMAS exhibited an eminent sensitivity to the frequency of sound. As seen in Figure 5G, the
multiresonant displacements of the PMAS membrane varied distinctly with frequencies from 100 Hz to
4 kHz, as measured by a laser Doppler vibrometer (LDV). On this basis, the multichannel data obtained by
PMAS embedded into the smartphone were transmitted to the processor, where the input signals were
compared with information in the built-in database with the help of the Gaussian mixture model algorithm
to complete the machine learning-based biometric verification, thus allowing or denying access to the
phone [Figure 5H]. The authors hypothesized that acoustic P-HMIs will not only play an important role in
clinical therapeutics, such as curing diseases and helping people with disabilities, but will also be used in
other areas, such as biometric authentication, due to the high sensitivity of piezoelectric devices to acoustic
waves. The promotion of acoustic therapeutic HMIs still faces many challenges, such as the miniaturization
of implanted devices competing with force-electric conversion efficiency and higher requirements for
biocompatibility and stability, which are critical issues that need to be addressed in the future.
Machine learning-enhanced P-HMIs
As a cutting-edge technology conducive to the enhancement of HMIs, machine learning has captured the
attention of researchers from various disciplines, especially those focusing on smart flexible electronics.
Incorporated with machine learning algorithms, the data obtained from flexible piezoelectric sensing
systems can be better utilized to provide appropriate guidance for a plethora of applications, including, but
not limited to, emotional detection[86,87], interactive displays[30,88,89] and sign-to-speech translation[90].

Machine learning not only improves the accuracy of results while reducing the volume of sensory
information that needs to be obtained, but also uncovers the hidden features in the data. Generally, there
are three types of machine learning models: supervised learning, unsupervised learning and reinforcement
learning, all of which can be utilized to enhance P-HMIs and boost the accuracy and efficiency of their
applications[57]. In particular, for stimuli identification and feature analysis in HMI devices, which require
training with real data, supervised learning algorithms are the most popular because they ensure fast and
efficient performance. For example, Lee et al. focused on the integration of machine learning with flexible
piezoelectric acoustic sensors, showing huge potential for speech recognition[91]. Figure 6A briefly illustrates
the process of machine learning-enhanced P-HMIs and their potential applications in HMIs concerning
human voices. The incorporation of novel piezoelectric acoustic sensors and optimized machine learning
algorithms is regarded as a key role of voice-controlling HMIs, such as biometric authentication, AI
secretaries and the control of smart home appliances.
Page 12 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

Figure 6. Machine learning-enhanced P-HMIs. (A) Overall illustration of process and potential applications of machine learning-
enhanced piezoelectric acoustic sensors. Reproduced with permission from ref.[91]. Copyright 2019, Wiley. (B) Visualized speaker
embeddings with t-SNE. Each speaker is marked by its color and the text indicates pronounced digits. Reproduced with permission
from ref.[92]. Copyright 2018, IEEE. (C) Artificial skin consisting of sensing, perception and encapsulation layers. Reproduced with
permission from ref.[93]. Copyright 2020, Wiley. (D) Architecture of neural network model for the classifier used for object recognition.
Reproduced with permission from ref.[94]. Copyright 2020, IOP Publishing Ltd. (E) Schematic diagram of piezoelectric output voltage
signals when a password is entered. (F) Image of piezoelectric panel for user authentication. Reproduced with permission from ref.[95].
Copyright 2021, IEEE. (G) Schematic illustration of HMIs with traditional and closed-loop HMI systems. Reproduced with permission
from ref.[97]. Copyright 2022, American Association for the Advancement of Science.

Novoselov et al. analyzed the performance of a speaker recognition system based on convolutional neural
networks (CNNs)[92]. A single high-level feature extractor was trained with human voices through deep
CNNs to distinguish speakers and digits simultaneously. With five randomly selected speakers, the same
digit pronounced by different people or different digits pronounced by the same person were projected on
two principal axes with t-distributed stochastic neighbor embedding (t-SNE), as displayed in Figure 6B. The
developed CNN system with multitask learning surpassed the classical algorithms, achieving an equal error
rate of 2.85%.

Furthermore, the combination of machine learning and e-skin has resulted in many interesting applications.
As depicted in Figure 6C, artificial skin usually consists of three layers: a sensing layer, a perception layer
and an encapsulation layer[93]. Combined with machine learning, Chung et al. successfully achieved object
classification with a piezoelectric actuator-sensor pair[94]. When the robotic hand grasped the objects, the
data obtained by the actuator-sensor pair was used as the input for machine learning after several
preprocessing steps. On this basis, a real-time object recognition system was successfully demonstrated with
a neural network classifier and the accuracy was above 90%. The classifier model for the object classification
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 13 of 19

task is schematically shown in Figure 6D, with the neural network at a scale of 1000 × 400 × 100 × 50 × 17.
This work successfully fulfilled object classification with flexible piezoelectric sensors and machine learning
algorithms, showing the significant potential of P-HMI systems for information exchange.

In addition to e-skin-based HMIs, machine learning can also be utilized for IoT devices, such as smart
interactive touch panels for door and mobile phone locks, making authentication more secure and reliable.
Researchers discovered that keystroke behavior varies significantly from person to person, an aspect that
was usually overlooked. Herein, Jia et al. demonstrated a keystroke authentication system with a
piezoelectric touch panel coordinated with a machine learning algorithm[95]. When a password was entered,
the keystroke datasets with force and time features could be obtained by the piezoelectric touch panel
[Figure 6E], which contains a 4 × 4 electrode array, as shown in Figure 6F. The machine learning algorithms
were then used to perform the authentication comparison, where the random forest classifier demonstrated
an accuracy of 96.40% due to the differences in individual keystroke behavior, which prevent intruders from
passing the authentication even if they enter the correct password.

With the rapid development of flexible electronics and machine learning, increasingly more applications
that once seemed impossible, such as brain-computer interactions, will become a reality. In recent studies,
soft piezoelectric sensors have been implanted into rodent nerves to detect nervous impulses[96]. With the
aid of machine learning, the obtained data can be effectively interpreted and then targeted instructions can
be issued. The study of machine learning-enhanced P-HMIs is still currently a budding research area.
Among all the machine learning algorithms, neural networks (NNs), with high robustness and fault
tolerance, have been the most popular in recent studies on P-HMIs, such as acoustic signal processing[91]. By
constructing a closed-loop HMI with both precise sensing and feedback [Figure 6G], it is expected that
people with disabilities, such as paralysis, will be able to control wheelchairs, TVs and even computers by
means other than their hands, and even convey thoughts for people with diseases, such as amyotrophic
lateral sclerosis, which will greatly improve their quality of life[97].

SUMMARY AND OUTLOOK

Since the discovery of the piezoelectric effect, researchers globally have been inspired to find approaches to
improve the performace and enrich the applications of piezoelectric devices. In this review, the recent
progress in P-HMIs has been presented around the applications in robotic control, the IoT, sports coaching,
acoustic therapeutics and machine learning. Given the latest trend in academic research on P-HMIs, we
have depicted a comprehensive diagram to explicate the summary and perspectives of P-HMI[98-103], as
shown in Figure 7. As the cornerstone of piezoelectric devices, piezoelectric materials largely determine
their comprehensive performance, such as sensitivity and response range. How to improve the
electromechanical coupling efficiency of piezoelectric materials, including organic materials, inorganic
materials, composite materials and biological materials, is the focus of future research, especially
considering both mechanical flexibility and piezoelectric output[98,100,104-110]. In addition, the effective design of
device structures also plays an important role in P-HMIs. Planar, sandwiched and three-dimensional
structures are three of the most popular architectures for HMIs[101-103,111]. The development of high-
performance piezoelectric devices by improving structural designs for scenario-specific application
requirements is another important research direction.

With the rapid development of various advanced technologies, such as VR/AR, cloud computing, 3D
printing and so on, the incorporation of P-HMIs with these technologies will be the focus of both academia
and industry. For instance, with VR/AR integrated P-HMIs, people will be able to chat with their friends,
watch movies or play video games more interactively and immerse themselves in the virtual world[40,41,44].
Page 14 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

Figure 7. Outlook for P-HMIs: from piezoelectric materials to future applications. Reproduced with permission from ref.[98]. Copyright
2008, Elsevier. Reproduced with permission from ref.[99]. Copyright 2009, AIP Publishing LLC. Reproduced with permission
from ref.[100]. Copyright 2020, Elsevier. Reproduced with permission from ref.[101]. Copyright 2019, Elsevier. Reproduced with permission
from ref.[102]. Copyright 2018, Wiley. Reproduced with permission from ref.[103]. Copyright 2019, Springer Nature.

With the development of cloud computing and big data, wearable P-HMIs can collect human health-related
physiological and behavioral information. On this basis, massive amounts of data would be processed and
computed on cloud platforms, which could extract more detailed information. Moreover, with high-
efficient data analysis on the cloud, applications such as remote smart medical care would be better
facilitated in the future[112,113]. Furthermore, benefiting from cutting-edge nanotechnology and tissue
engineering techniques, P-HMI will become even more miniaturized and friendly for human bodies.
Combined with machine learning and other algorithms, P-HMI with large-scale sensing and feedback
shows enormous promise in many fields, such as intelligent diagnostics, smart education, biometrics and so
on, which is of great significance for improving the quality and convenience of our daily lives. However,
there are still many challenges that must be addressed for P-HMIs to fulfill the above-mentioned smart
applications:

(1) The competitive relationship between mechanical flexibility and piezoelectric output has been the focus
of research on flexible piezoelectric devices for a long time. How to balance their relationship to obtain high
performance for P-HMI needs to be further explored, especially considering the piezoelectric effect for
sensing and the inverse piezoelectric effect for actuation in an integrated manner.

(2) Many different working principles can be used for sensing and actuation of HMIs and the fusion of
multiple principles is an important development trend. The coupling of different principles expects hybrid
flexible HMI devices to exhibit better sensitivity, wider detection range and higher signal-to-noise ratio.

(3) To obtain more accurate and higher-resolution monitoring information, the number of sensors used for
HMIs will continue to increase. Along with this, it poses higher challenges for the large-scale preparation of
piezoelectric devices and the processing of massive amounts of information.
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 15 of 19

(4) P-HMIs usually involve multidisciplinary intersections, including materials science, mechanical
engineering, electronic technology, biomedical engineering and so on. In the future, scientists and engineers
will need more interdisciplinary knowledge to engage in the complicated R&D processes of P-HMIs to
advance their practical applications.

In summary, flexible piezoelectric human-machine interactive devices have sparked numerous ideas and
thoughts in both academia and industry. Although the fantasies in science fiction would not come true all at
once, increasingly more meaningful applications will be realized with the development of P-HMIs.
Intelligent HMIs will bring convenience and efficiency to all aspects of our daily lives, raising them to a level
we could never have imagined.

DECLARATIONS
Authors’ contributions
Conceptualization, investigation, wrote the original draft: Liu J
Reviewed and revised the manuscript: Tian G, Deng W
Supervision, project administration, funding acquisition: Yang W, Deng W
Availability of data and materials
Not applicable.
Financial support and sponsorship
This research was financially supported by the National Natural Science Foundation of China
(No.61801403) and the Basic Research Cultivation Project (No.2682021ZTPY004).
Conflicts of interest
All authors declared that there are no conflicts of interest.
Ethical approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Copyright
© The Author(s) 2022.

REFERENCES
1. Zhu M, He T, Lee C. Technologies toward next generation human machine interfaces: From machine learning enhanced tactile
sensing to neuromorphic sensory systems. Appl Phys Rev 2020;7:031305. DOI
2. Cao X, Xiong Y, Sun J, Zhu X, Sun Q, Wang ZL. Piezoelectric nanogenerators derived self-powered sensors for multifunctional
applications and Artificial Intelligence. Adv Funct Mater 2021;31:2102983. DOI
3. Hua Q, Cui X, Ji K, Wang B, Hu W. Piezotronics enabled artificial intelligence systems. J Phys Mater 2021;4:022003. DOI
4. Yin R, Wang D, Zhao S, Lou Z, Shen G. Wearable sensors-enabled human machine interaction systems: from design to application.
Adv Funct Mater 2021;31:2008936. DOI
5. Araromi OA, Graule MA, Dorsey KL, et al. Ultra-sensitive and resilient compliant strain gauges for soft machines. Nature
2020;587:219-24. DOI PubMed
6. Lee S, Shi Q, Lee C. From flexible electronics technology in the era of IoT and artificial intelligence toward future implanted body
sensor networks. APL Mater 2019;7:031302. DOI
7. Ma M, Zhang Z, Liao Q, et al. Self-powered artificial electronic skin for high-resolution pressure sensing. Nano Energy 2017;32:389-
96. DOI
Page 16 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

8. Yi Z, Liu Z, Li W, et al. Piezoelectric dynamics of arterial pulse for wearable continuous blood pressure monitoring. Adv Mater
2022;34:e2110291. DOI PubMed
9. Zheng Q, Shi B, Li Z, Wang ZL. Recent progress on piezoelectric and triboelectric energy harvesters in biomedical systems. Adv Sci
2017;4:1700029. DOI PubMed PMC
10. Park DY, Joe DJ, Kim DH, et al. Self-powered real-time arterial pulse monitoring using ultrathin epidermal piezoelectric sensors. Adv
Mater 2017;29:1702308. DOI PubMed
11. Rao Z, Ershad F, Almasri A, Gonzalez L, Wu X, Yu C. Soft electronics for the skin: from health monitors to human-machine
interfaces. Adv Mater Technol 2020;5:2000233. DOI
12. Kim N, Lee JM, Moradnia M, et al. Biocompatible composite thin-film wearable piezoelectric pressure sensor for monitoring of
physiological and muscle motions. Soft Sci 2022;2:8. DOI
13. Patel S, Ershad F, Zhao M, et al. Wearable electronics for skin wound monitoring and healing. Soft Sci 2022;2:9. DOI
14. Yan C, Deng W, Jin L, et al. Epidermis-inspired ultrathin 3D cellular sensor array for self-powered biomedical monitoring. ACS Appl
Mater Interfaces 2018;10:41070-5. DOI PubMed
15. Ozioko O, Dahiya R. Smart tactile gloves for haptic interaction, communication, and rehabilitation. Adv Intell Syst 2022;4:2100091.
DOI
16. Gao C, Long Z, Zhong T, Liang S, Xing L. A self-powered intelligent glove for real-time human-machine gesture interaction based
on piezoelectric effect of T-ZnO/PVDF film. J Phys D Appl Phys 2022;55:194004. DOI
17. Fuh YK, Wang BS. Near field sequentially electrospun three-dimensional piezoelectric fibers arrays for self-powered sensors of
human gesture recognition. Nano Energy 2016;30:677-83. DOI
18. Deng C, Tang W, Liu L, Chen B, Li M, Wang ZL. Self-powered insole plantar pressure mapping system. Adv Funct Mater
2018;28:1801606. DOI
19. Zhu M, Shi Q, He T, et al. Self-powered and self-functional cotton sock using piezoelectric and triboelectric hybrid mechanism for
healthcare and sports monitoring. ACS Nano 2019;13:1940-52. DOI PubMed
20. Deng W, Yang T, Jin L, et al. Cowpea-structured PVDF/ZnO nanofibers based flexible self-powered piezoelectric bending motion
sensor towards remote control of gestures. Nano Energy 2019;55:516-25. DOI
21. Rao J, Chen Z, Zhao D, Yin Y, Wang X, Yi F. Recent progress in self-powered skin sensors. Sensors 2019;19:2763. DOI PubMed
PMC
22. Chen Y, Gao Z, Zhang F, Wen Z, Sun X. Recent progress in self-powered multifunctional e-skin for advanced applications.
Exploration 2022;2:20210112. DOI
23. Lee S, Hinchet R, Lee Y, et al. Ultrathin nanogenerators as self-powered/active skin sensors for tracking eye ball motion. Adv Funct
Mater 2014;24:1163-8. DOI
24. Cha Y, Seo J, Kim J, Park J. Human-computer interface glove using flexible piezoelectric sensors. Smart Mater Struct
2017;26:057002. DOI
25. Wen F, Sun Z, He T, et al. Machine learning glove using self-powered conductive superhydrophobic triboelectric textile for gesture
recognition in VR/AR applications. Adv Sci 2020;7:2000261. DOI PubMed PMC
26. Wang C, Fan Z, Feng K. A self-power flexible piezoelectric sensing system for badminton training monitoring. IEICE Electron Expr
2021;18:20210119-20210119. DOI
27. Zhou P, Zheng Z, Wang B, Guo Y. Self-powered flexible piezoelectric sensors based on self-assembled 10 nm BaTiO3 nanocubes on
glass fiber fabric. Nano Energy 2022;99:107400. DOI
28. Liu H, Dong W, Li Y, et al. An epidermal sEMG tattoo-like patch as a new human-machine interface for patients with loss of voice.
Microsyst Nanoeng 2020;6:16. DOI PubMed PMC
29. Reid T, Gibert J. Inclusion in human-machine interactions. Science 2022;375:149-50. DOI PubMed
30. Gao S, Duan J, Kitsos V, Selviah DR, Nathan A. User-oriented piezoelectric force sensing and artificial neural networks in interactive
displays. IEEE J Electron Devices Soc 2018;6:766-73. DOI
31. Zhang H, Tian G, Xiong D, et al. Carrier concentration-dependent interface engineering for high-performance zinc oxide
piezoelectric device. J Colloid Interface Sci 2023;629:534-40. DOI PubMed
32. Gao Y, Yan C, Huang H, et al. Microchannel-confined MXene based flexible piezoresistive multifunctional micro-force sensor. Adv
Funct Mater 2020;30:1909603. DOI
33. Li J, Fang L, Sun B, et al. Recent progress in flexible and stretchable piezoresistive sensors and their applications. J Electrochem Soc
2020;167:037561. DOI
34. Li J, Fang L, Sun B, Li X, Kang SH. Fingerprint-enhanced capacitive-piezoelectric flexible sensing skin to discriminate static and
dynamic tactile stimuli. Adv Intell Syst 2019;1:1900051. DOI
35. Cotton DPJ, Graz IM, Lacour SP. A multifunctional capacitive sensor for stretchable electronic skins. IEEE Sensors J 2009;9:2008-9.
DOI
36. Dong K, Peng X, Wang ZL. Fiber/fabric-based piezoelectric and triboelectric nanogenerators for flexible/stretchable and wearable
electronics and Artificial Intelligence. Adv Mater 2020;32:e1902549. DOI PubMed
37. Zhu J, Zhou C, Zhang M. Recent progress in flexible tactile sensor systems: from design to application. Soft Sci 2021;1:3. DOI
38. Zhang Z, Wen F, Sun Z, Guo X, He T, Lee C. Artificial intelligence-enabled sensing technologies in the 5G/internet of things era:
from virtual reality/augmented reality to the digital twin. Adv Intell Syst 2022;4:2100228. DOI
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 17 of 19

39. Wang W, Tian Y, Wang X, et al. Ethanol sensing properties of porous ZnO spheres via hydrothermal route. J Mater Sci
2013;48:3232-8. DOI
40. Song K, Kim SH, Jin S, et al. Pneumatic actuator and flexible piezoelectric sensor for soft virtual reality glove system. Sci Rep
2019;9:8988. DOI PubMed PMC
41. Yu X, Xie Z, Yu Y, et al. Skin-integrated wireless haptic interfaces for virtual and augmented reality. Nature 2019;575:473-9. DOI
PubMed
42. Hwang G, Kim Y, Lee J, et al. Self-powered deep brain stimulation via a flexible PIMNT energy harvester. Energy Environ Sci
2015;8:2677-84. DOI
43. Qiu Y, Tian Y, Sun S, et al. Bioinspired, multifunctional dual-mode pressure sensors as electronic skin for decoding complex loading
processes and human motions. Nano Energy 2020;78:105337. DOI
44. Meier M, Streli P, Fender A, et al. TapID: rapid touch interaction in virtual reality using wearable sensing. IEEE VR 2021:519-528.
DOI
45. Lu Z, Zhu Y, Jia C, et al. A self-powered portable flexible sensor of monitoring speed skating techniques. Biosensors 2021;11:108.
DOI PubMed PMC
46. Han JH, Bae KM, Hong SK, et al. Machine learning-based self-powered acoustic sensor for speaker recognition. Nano Energy
2018;53:658-65. DOI
47. Sun Z, Zhu M, Chen Z, et al. Haptic-feedback ring enabled human-machine interface (HMI) aiming at immersive virtual reality
experience. In 21st International Conference on Solid-State Sensors, Actuators and Microsystems (Transducers); 2021. pp. 333-6.
DOI
48. Zhu G, Wang AC, Liu Y, Zhou Y, Wang ZL. Functional electrical stimulation by nanogenerator with 58 V output voltage. Nano Lett
2012;12:3086-90. DOI PubMed
49. Charalambides A, Bergbreiter S. Rapid manufacturing of mechanoreceptive skins for slip detection in robotic grasping. Adv Mater
Technol 2017;2:1600188. DOI
50. Chhowalla M, Shin HS, Eda G, Li LJ, Loh KP, Zhang H. The chemistry of two-dimensional layered transition metal dichalcogenide
nanosheets. Nat Chem 2013;5:263-75. DOI PubMed
51. Yang P, Chou S, Hsu C, et al. Tin disulfide piezoelectric nanogenerators for biomechanical energy harvesting and intelligent human-
robot interface applications. Nano Energy 2020;75:104879. DOI
52. Ilyas MA, Swingler J. Piezoelectric energy harvesting from raindrop impacts. Energy 2015;90:796-806. DOI
53. Ahn JH, Lee MJ, Heo H, et al. Deterministic two-dimensional polymorphism growth of hexagonal n-type SnS2 and orthorhombic p-
type SnS crystals. Nano Lett 2015;15:3703-8. DOI PubMed
54. Lim S, Son D, Kim J, et al. Transparent and stretchable interactive human machine interface based on patterned graphene
heterostructures. Adv Funct Mater 2015;25:375-83. DOI
55. Lin W, Wang B, Peng G, Shan Y, Hu H, Yang Z. Skin-inspired piezoelectric tactile sensor array with crosstalk-free row+column
electrodes for spatiotemporally distinguishing diverse stimuli. Adv Sci 2021;8:2002817. DOI PubMed PMC
56. Kim K, Kim J, Choi J, Kim J, Lee S. Depth camera-based 3D hand gesture controls with immersive tactile feedback for natural mid-
air gesture interactions. Sensors 2015;15:1022-46. DOI PubMed PMC
57. Heng W, Solomon S, Gao W. Flexible electronics and devices as human-machine interfaces for medical robotics. Adv Mater
2022;34:e2107902. DOI PubMed PMC
58. Haroun A, Le X, Gao S, et al. Progress in micro/nano sensors and nanoenergy for future AIoT-based smart home applications. Nano
Express 2021;2:022005. DOI
59. Dong B, Shi Q, Yang Y, Wen F, Zhang Z, Lee C. Technology evolution from self-powered sensors to AIoT enabled smart homes.
Nano Energy 2021;79:105414. DOI
60. Lee WH, Chiu CY. Design and implementation of a smart traffic signal control system for smart city applications. Sensors
2020;20:508. DOI PubMed PMC
61. Gao W, Emaminejad S, Nyein HYY, et al. Fully integrated wearable sensor arrays for multiplexed in situ perspiration analysis.
Nature 2016;529:509-14. DOI PubMed PMC
62. Dai Y, Chen J, Tian W, Xu L, Gao S. A PVDF/Au/PEN multifunctional flexible human-machine interface for multidimensional
sensing and energy harvesting for the internet of things. IEEE Sensors J 2020;20:7556-68. DOI
63. Song GJ, Cho JY, Kim K, et al. Development of a pavement block piezoelectric energy harvester for self-powered walkway
applications. Appl Energy 2019;256:113916. DOI
64. Kim JH, Cho JY, Jhun JP, et al. Development of a hybrid type smart pen piezoelectric energy harvester for an IoT platform. Energy
2021;222:119845. DOI
65. Lv P, Qian J, Yang C, et al. Flexible all-inorganic Sm-doped PMN-PT film with ultrahigh piezoelectric coefficient for mechanical
energy harvesting, motion sensing, and human-machine interaction. Nano Energy 2022;97:107182. DOI
66. Su Y, Li W, Yuan L, et al. Piezoelectric fiber composites with polydopamine interfacial layer for self-powered wearable
biomonitoring. Nano Energy 2021;89:106321. DOI
67. Kim K, Cho JY, Jabbar H, et al. Optimized composite piezoelectric energy harvesting floor tile for smart home energy management.
Energy Convers Manag 2018;171:31-7. DOI
68. Le X, Shi Q, Vachon P, Ng EJ, Lee C. Piezoelectric MEMS-evolution from sensing technology to diversified applications in the
Page 18 of 19 Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23

5G/Internet of Things (IoT) era. J Micromech Microeng 2022;32:014005. DOI

69. Liu L, Guo X, Liu W, Lee C. Recent progress in the energy harvesting technology-from self-powered sensors to self-sustained IoT,
and new applications. Nanomaterials 2021;11:2975. DOI
70. Jordan MI, Mitchell TM. Machine learning: trends, perspectives, and prospects. Science 2015;349:255-60. DOI
71. Jian G, Jiao Y, Meng Q, Shao H, Wang F, Wei Z. 3D BaTiO3 flower based polymer composites exhibiting excellent piezoelectric
energy harvesting properties. Adv Mater Interfaces 2020;7:2000484. DOI
72. Zhao C, Jia C, Zhu Y, Zhao T. An effective self-powered piezoelectric sensor for monitoring basketball skills. Sensors 2021;21:5144.
DOI
73. Liu W, Long Z, Yang G, Xing L. A self-powered wearable motion sensor for monitoring volleyball skill and building big sports data.
Biosensors 2022;12:60. DOI
74. Tian G, Deng W, Gao Y, et al. Rich lamellar crystal baklava-structured PZT/PVDF piezoelectric sensor toward individual table tennis
training. Nano Energy 2019;59:574-81. DOI
75. Yao D, Cui H, Hensleigh R, et al. Achieving the upper bound of piezoelectric response in tunable, wearable 3D printed
nanocomposites. Adv Funct Mater 2019;29:1903866. DOI
76. Mao Y, Zhu Y, Zhao T, et al. A portable and flexible self-powered multifunctional sensor for real-time monitoring in swimming.
Biosensors 2021;11:147. DOI PubMed PMC
77. Han JH, Kwak J, Joe DJ, et al. Basilar membrane-inspired self-powered acoustic sensor enabled by highly sensitive multi tunable
frequency band. Nano Energy 2018;53:198-205. DOI
78. Viola G, Chang J, Maltby T, et al. Bioinspired multiresonant acoustic devices based on electrospun piezoelectric polymeric
nanofibers. ACS Appl Mater Interfaces 2020;12:34643-57. DOI PubMed PMC
79. Wang J, He T, Lee C. Development of neural interfaces and energy harvesters towards self-powered implantable systems for
healthcare monitoring and rehabilitation purposes. Nano Energy 2019;65:104039. DOI
80. Shi Q, Wang T, Lee C. MEMS based broadband piezoelectric ultrasonic energy harvester (PUEH) for enabling self-powered
implantable biomedical devices. Sci Rep 2016;6:24946. DOI PubMed PMC
81. Jiang L, Yang Y, Chen R, et al. Ultrasound-induced wireless energy harvesting for potential retinal electrical stimulation application.
Adv Funct Mater 2019;29:1902522. DOI
82. Zhang T, Liang H, Wang Z, et al. Piezoelectric ultrasound energy-harvesting device for deep brain stimulation and analgesia
applications. Sci Adv 2022;8:eabk0159. DOI PubMed PMC
83. Jiang L, Lu G, Zeng Y, et al. Flexible ultrasound-induced retinal stimulating piezo-arrays for biomimetic visual prostheses. Nat
Commun 2022;13:3853. DOI PubMed PMC
84. Piech DK, Johnson BC, Shen K, et al. A wireless millimetre-scale implantable neural stimulator with ultrasonically powered
bidirectional communication. Nat Biomed Eng 2020;4:207-22. DOI PubMed
85. Wang HS, Hong SK, Han JH, et al. Biomimetic and flexible piezoelectric mobile acoustic sensors with multiresonant ultrathin
structures for machine learning biometrics. Sci Adv 2021;7:eabe5683. DOI PubMed PMC
86. Kyamakya K, Al-Machot F, Haj Mosa A, Bouchachia H, Chedjou JC, Bagula A. Emotion and stress recognition related sensors and
machine learning technologies. Sensors 2021;21:2273. DOI PubMed PMC
87. Qi Y, Jia W, Gao S. Emotion recognition based on piezoelectric keystroke dynamics and Machine Learning. In 2021 IEEE
International Conference on Flexible and Printable Sensors and Systems (FLEPS); 2021. pp. 1-4. DOI
88. Gao S, Dai Y, Kitsos V, Wan B, Qu X. High three-dimensional detection accuracy in piezoelectric-based touch panel in interactive
displays by optimized artificial neural networks. Sensors 2019;19:753. DOI PubMed PMC
89. Gao S, Guo R, Shao M, Xu L. A touch orientation classification-based force-voltage responsivity stabilization method for
piezoelectric force sensing in interactive displays. IEEE Sensors J 2020;20:8147-54. DOI
90. Zhou Z, Chen K, Li X, et al. Sign-to-speech translation using machine-learning-assisted stretchable sensor arrays. Nat Electron
2020;3:571-8. DOI
91. Jung YH, Hong SK, Wang HS, et al. Flexible piezoelectric acoustic sensors and machine learning for speech processing. Adv Mater
2020;32:e1904020. DOI PubMed
92. Novoselov S, Kudashev O, Shchemelinin V, et al. Deep CNN based feature extractor for text-prompted speaker recognition. In 2018
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP); 2018. pp. 5334-8. DOI
93. Wang M, Luo Y, Wang T, et al. artificial skin perception. Adv Mater 2021;33:e2003014. DOI
94. Chung J, Lim H, Lim M, Cha Y. Object classification based on piezoelectric actuator-sensor pair on robot hand using neural network.
Smart Mater Struct 2020;29:105020. DOI
95. Jia W, Qi Y, Huang A, Zhou F, Gao S. High security user authentication based on piezoelectric keystroke dynamics applying to
multiple emotional responses. IEEE Sensors J 2022;22:2814-22. DOI
96. Jeong YC, Lee HE, Shin A, Kim DG, Lee KJ, Kim D. Progress in brain-compatible interfaces with soft nanomaterials. Adv Mater
2020;32:e1907522. DOI PubMed
97. Liu Y, Yiu C, Song Z, et al. Electronic skin as wireless human-machine interfaces for robotic VR. Sci Adv 2022;8:eabl6700. DOI
PubMed PMC
98. Zhu G, Zeng Z, Zhang L, et al. Piezoelectricity in β-phase PVDF crystals: a molecular simulation study. Comput Mater Sci
2008;44:224-9. DOI
Liu et al. Soft Sci 2022;2:22 https://dx.doi.org/10.20517/ss.2022.23 Page 19 of 19

99. Zhang J, Wang C, Bowen C. Piezoelectric effects and electromechanical theories at the nanoscale. Nanoscale 2014;6:13314-27. DOI
100. Zhang M, Zhang AM, Chen Y, et al. Polyoxovanadate-polymer hybrid electrolyte in solid state batteries. Energy Stor Mater
2020;29:172-81. DOI
101. Zhang L, Gui J, Wu Z, et al. Enhanced performance of piezoelectric nanogenerator based on aligned nanofibers and three-
dimensional interdigital electrodes. Nano Energy 2019;65:103924. DOI
102. Chu Y, Zhong J, Liu H, et al. Human pulse diagnosis for medical assessments using a wearable piezoelectret sensing system. Adv
Funct Mater 2018;28:1803413. DOI
103. Han M, Wang H, Yang Y, et al. Three-dimensional piezoelectric polymer microsystems for vibrational energy harvesting, robotic
interfaces and biomedical implants. Nat Electron 2019;2:26-35. DOI
104. Liu X, Wei Y, Qiu Y. Advanced flexible skin-like pressure and strain sensors for human health monitoring. Micromachines
2021;12:695. DOI
105. Jella V, Ippili S, Eom J, et al. A comprehensive review of flexible piezoelectric generators based on organic-inorganic metal halide
perovskites. Nano Energy 2019;57:74-93. DOI
106. Liu H, Zhong J, Lee C, Lee S, Lin L. A comprehensive review on piezoelectric energy harvesting technology: materials, mechanisms,
and applications. Appl Phys Rev 2018;5:041306. DOI
107. Kang S, Kim SH, Lee HB, et al. High-power energy harvesting and imperceptible pulse sensing through peapod-inspired
hierarchically designed piezoelectric nanofibers. Nano Energy 2022;99:107386. DOI
108. Mahapatra SD, Mohapatra PC, Aria AI, et al. Piezoelectric materials for energy harvesting and sensing applications: roadmap for
future smart materials. Adv Sci 2021;8:e2100864. DOI
109. Yi F, Ren H, Shan J, Sun X, Wei D, Liu Z. Wearable energy sources based on 2D materials. Chem Soc Rev 2018;47:3152-88. DOI
110. Gao Z, Zhou J, Gu Y, et al. Effects of piezoelectric potential on the transport characteristics of metal-ZnO nanowire-metal field effect
transistor. J Appl Phys 2009;105:113707. DOI
111. Huang X, Wang Y, Zhang X. Ultrarobust, hierarchically anisotropic structured piezoelectric nanogenerators for self-powered sensing.
Nano Energy 2022;99:107379. DOI
112. Chen G, Xiao X, Zhao X, Tat T, Bick M, Chen J. Electronic textiles for wearable point-of-care systems. Chem Rev 2022;122:3259-
91. DOI
113. Deng W, Zhou Y, Libanori A, Chen G, Yang W, Chen J. Piezoelectric nanogenerators for personalized healthcare. Chem Soc Rev
2022;51:3380-435. DOI PubMed