clear close all clc %% 参数设置 N = 612; % 子载波数量 cp_len = 40; % 循环前缀长度(通常取1/4~1/8符号长度) num_symbols = 14; % OFDM符号数量 mod_order = 16; % 调制阶数(4=QPSK, 16=16QAM...) Fs = 122.88e6; % 采样率 20MHz (典型LTE参数) Ts = 1/Fs; % 采样间隔 %% 1. 生成随机数据 num_bits = num_symbols*N*log2(mod_order); tx_bits = randi([0 1], num_bits, 1); %% 2. QAM调制 tx_symbols = qammod(tx_bits, mod_order, 'InputType', 'bit', 'UnitAveragePower', true); %% 3. 串并转换 tx_symbols_matrix = reshape(tx_symbols, N, num_symbols); %% 4. IFFT变换 tx_time = ifft(tx_symbols_matrix, N); %% 5. 添加循环前缀 tx_with_cp = [tx_time(end-cp_len+1:end, :); tx_time]; %% 6. 并串转换 ofdm_signal = tx_with_cp(:); % 转换为列向量 y=ofdm_signal;%输入信号 h=0 0.0417241926862380 -0.224836932944330 1.27671208956916 1.56109585209425 0.180360835931948 -0.0196317923804101 0.0122204462346232 0.0185923956511133 0.216396514812836 0.0640418001415771 0.109989218142588 0.0124396997597218 0.0648767636295433 0.0169102921612485 -0.00158507654306856 -0.000789181453279766 0.0490257552169616 0.0225516430581802 0.0247939397873611 0.0222928886341433 0.0152144577137114 0.00461117193825012 -0.000418008529206487 0 0 0 0 0 0 0 0 0 0 0 2.38558151107684e-05 -0.000155008019853569 0.000616514024510853 0.000630763837128181 -4.46063516586269e-05; y1=conv(y,h,"same");%y为输入信号 N1 = length(y1); % 信号长度，假设 x 是输入信号 % % 修正后的瑞利衰落生成（频域实现） % H_freq = (randn(N,1) + 1i*randn(N,1))/sqrt(2); % 频域信道响应 % H_time = ifft(H_freq); % 时域信道冲激响应 % % % 通过线性卷积模拟多径效应（时域实现） % h1 = H_time(1:cp_len); % 取前cp_len个抽头保证循环前缀有效性 % x_faded = conv(y, h1); % 多径信道传输 % 生成瑞利衰落（两个独立的高斯分布） % real_part = randn(1, N1); % 实部，高斯噪声 % imag_part = randn(1, N1); % 虚部，高斯噪声 % % % 组合成复数信号 % rayleigh_fading = sqrt(real_part.^2 + imag_part.^2); % 瑞利衰落幅度 % % % 将瑞利衰落应用到信号上 % x_faded = y1 .* rayleigh_fading; rayleigh_fading = (randn(N1, 1) + 1i*randn(N1, 1)) / sqrt(2); % 复数瑞利衰落 x_faded = y1 .* rayleigh_fading; % 应用复数衰落 % 绘制衰落前后的信号对比（如果 x 是一个实信号） figure; subplot(2, 1, 1); plot(real(y1)); % 原始信号的实部 title('Original Signal'); subplot(2, 1, 2); plot(real( x_faded)); % 加入瑞利衰落后的实部 title('Signal after Rayleigh Fading'); %% 接收信号处理 rx_signal = x_faded(1:length(ofdm_signal)); % 截断卷积带来的延拓 % valid_length = floor(length(rx_signal)/(N + cp_len)) * (N + cp_len); % rx_signal = rx_signal(1:valid_length); % 截断至合法长度 % 串并转换 rx_with_cp = reshape(rx_signal, N + cp_len, []); % 去除循环前缀 rx_time = rx_with_cp(cp_len+1:end, :); % FFT变换到频域 rx_freq = fft(rx_time, N); figure; stem(abs(h)); title('信道冲激响应幅度'); xlabel('抽头序号'); %% 并串转换与QAM解调 rx_symbols = rx_freq(:); % 展开为列向量 % QAM解调 rx_bits = qamdemod(rx_symbols, mod_order, 'OutputType', 'bit', 'UnitAveragePower', true); %% 误码率计算 ber = sum(rx_bits ~= tx_bits) / num_bits; fprintf('误码率: %.4f\n', ber); load("trainedChannelEstimationNetwork.mat") % Concatenate the real and imaginary grids along the batch dimension nnInput = cat(4,real(rx_freq),imag(rx_freq)); % Use the neural network to estimate the channel if canUseGPU nnInput = gpuArray(nnInput); end estChannelGridNN = predict(channelEstimationCNN,nnInput); % Convert results to complex estChannelGridNN = complex(estChannelGridNN(:,:,:,1),estChannelGridNN(:,:,:,2)); %% % Calculate the mean squared error (MSE) of each estimation method. %% 接收端信道估计 % H_est = rx_freq(:,1) ./ pilot_symbols; % 使用导频估计信道响应 %% 均衡所有符号（迫零均衡） rx_eq = rx_freq ./ estChannelGridNN; %% 并串转换与QAM解调 rx_symbols1 = rx_eq(:); % 展开为列向量 % QAM解调 rx_bits1 = qamdemod(rx_symbols1, mod_order, 'OutputType', 'bit', 'UnitAveragePower', true); %% 误码率计算 ber = sum(rx_bits1 ~= tx_bits) / num_bits; fprintf('均衡后误码率: %.4f\n', ber);