c语言谱减法实现wav音频降噪

谱减法是一种常用的音频降噪方法，可以有效地降低噪声，提升音频质量。在 C 语言中，通过 FFT 算法将时域信号转换为频域信号，然后对每个频率分量的振幅进行计算，并相应地减去一定的阈值，从而实现音频的降噪。

下面是一个简单的 C 语言程序，用于实现基于谱减法的音频降噪：

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <string.h>

#define SAMPLERATE 44100
#define FRAME_SIZE 2048
#define HOP_SIZE 1024
#define THR_FACTOR 1.5

typedef struct {
    float re;
    float im;
} complex;

float get_threshold(float *power_spectrum, int length, float factor) {
    float sum = 0.0;
    for (int i = 0; i < length; i++) {
        sum += power_spectrum[i];
    }
    float mean = sum / length;
    return mean * factor;
}

void apply_spectrogram_threshold(float *re, float *im, int length, float threshold) {
    for (int i = 0; i < length; i++) {
        if (sqrt(re[i] * re[i] + im[i] * im[i]) <= threshold) {
            re[i] = 0;
            im[i] = 0;
        }
    }
}

void apply_ifft(float *re, float *im, int length, float *output) {
    complex *in = (complex *) malloc(length * sizeof(complex));
    complex *out = (complex *) malloc(length * sizeof(complex));
    for (int i = 0; i < length; i++) {
        in[i].re = re[i];
        in[i].im = im[i];
    }
    for (int i = 0; i < length; i++) {
        out[i].re = 0;
        out[i].im = 0;
        for (int j = 0; j < length; j++) {
            float angle = 2.0 * M_PI * i * j / length;
            out[i].re += in[j].re * cos(angle) - in[j].im * sin(angle);
            out[i].im += in[j].re * sin(angle) + in[j].im * cos(angle);
        }
        out[i].re /= length;
        out[i].im /= length;
    }
    for (int i = 0; i < length; i++) {
        output[i] = out[i].re;
    }
}

int main(int argc, char **argv) {
    if (argc != 3) {
        printf("Usage: %s input.wav output.wav\n", argv[0]);
        return 1;
    }
    
    char *input_file = argv[1];
    char *output_file = argv[2];
    
    FILE *input = fopen(input_file, "rb");
    if (!input) {
        printf("Error: Failed to open %s\n", input_file);
        return 1;
    }
    
    FILE *output = fopen(output_file, "wb");
    if (!output) {
        printf("Error: Failed to open %s\n", output_file);
        fclose(input);
        return 1;
    }
    
    fseek(input, 0, SEEK_END);
    int length = ftell(input);
    fseek(input, 0, SEEK_SET);
    
    float *buffer = (float *) malloc(length);
    fread(buffer, sizeof(float), length, input);
    
    fclose(input);
    
    // Spectrogram
    float *window = (float *) malloc(FRAME_SIZE * sizeof(float));
    for (int i = 0; i < FRAME_SIZE; i++) {
        window[i] = 0.5 - 0.5 * cos(2.0 * M_PI * i / FRAME_SIZE);
    }
    int num_frames = (length - FRAME_SIZE) / HOP_SIZE + 1;
    float *spectrum = (float *) malloc(FRAME_SIZE * sizeof(float));
    float *power_spectrum = (float *) malloc(FRAME_SIZE / 2 * sizeof(float));
    float *output_buffer = (float *) malloc(length);
    memset(output_buffer, 0, length);
    for (int i = 0; i < num_frames; i++) {
        memset(spectrum, 0, FRAME_SIZE * sizeof(float));
        memcpy(spectrum, &buffer[i * HOP_SIZE], FRAME_SIZE * sizeof(float));
        for (int j = 0; j < FRAME_SIZE; j++) {
            spectrum[j] *= window[j];
        }
        apply_ifft(spectrum, &spectrum[FRAME_SIZE / 2], FRAME_SIZE, output_buffer);
        for (int j = 0; j < FRAME_SIZE / 2; j++) {
            power_spectrum[j] = output_buffer[j] * output_buffer[j] + output_buffer[FRAME_SIZE / 2 + j] * output_buffer[FRAME_SIZE / 2 + j];
        }
        float threshold = get_threshold(power_spectrum, FRAME_SIZE / 2, THR_FACTOR);
        apply_spectrogram_threshold(spectrum, &spectrum[FRAME_SIZE / 2], FRAME_SIZE / 2, threshold);
        apply_ifft(spectrum, &spectrum[FRAME_SIZE / 2], FRAME_SIZE, output_buffer);
        for (int j = 0; j < FRAME_SIZE; j++) {
            output_buffer[j] *= window[j];
            output_buffer[j] *= 1.0 / HOP_SIZE;
            output_buffer[j] += buffer[i * HOP_SIZE + j];
        }
        memcpy(&output_buffer[i * HOP_SIZE], output_buffer, FRAME_SIZE * sizeof(float));
    }
    
    fwrite(output_buffer, sizeof(float), length, output);
    fclose(output);
    
    free(buffer);
    free(window);
    free(spectrum);
    free(power_spectrum);
    free(output_buffer);
    
    return 0;
}

该程序实现了对 WAV 格式的音频文件进行谱减法降噪，并将降噪后的结果保存到另一个 WAV 文件中。首先，程序读取输入文件，并将其中的浮点采样点存储在一个缓冲区中。然后，程序对缓冲区中的采样点进行谱减法处理，将每个帧的功率谱计算出来，并根据平均值和阈值系数计算出一个阈值，然后将小于阈值的功率谱设置为零，从而实现噪声的去除。最后，程序将降噪后的采样点存储到输出文件中。