Python 筛选有人声的Wav文件

用途：

判断左声道/右声道，是否有疑是人说话的声音，筛选有声音但SR无法识别出的wav文件，
过滤无效的声音，短促的声音等

Code:

# -*- coding: utf-8 -*-
# 读Wave文件并且绘制波形
import copy
import wave

import matplotlib.pyplot as plt
import numpy as np

plt.rcParams['font.sans-serif'] = ['SimHei']  # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False  # 用来正常显示符号


def ProcessWav(wavPath, effectiveScore=0.02, ignoreScore=0.03, validTime=200):
    """

    :param wavPath: 音频文件
    :param effectiveScore: 有声音的平均阈值
    :param ignoreScore: 无效声音的阈值
    :param validTime: 说话的有效时长：200ms
    :return:
    """
    try:
        f = wave.open(wavPath, "rb")

        # 读取格式信息
        # (声道数、量化位数、采样频率、采样点数、压缩类型、压缩类型的描述)
        # (nchannels, sampwidth, framerate, nframes, comptype, compname)
        params = f.getparams()
        nchannels, sampwidth, framerate, nframes = params[:4]
        # nchannels通道数 = 2
        # sampwidth量化位数 = 2
        # framerate采样频率 = 22050
        # nframes采样点数 = 53395

        # 读取nframes个数据，返回字符串格式
        str_data = f.readframes(nframes)

        f.close()
        # 将字符串转换为数组，得到一维的short类型的数组
        wave_data = np.frombuffer(str_data, dtype=np.short)

        # 赋值的归一化
        wave_data = wave_data * 1.0 / (max(abs(wave_data)))

        # 整合左声道和右声道的数据
        wave_data = np.reshape(wave_data, [nframes, nchannels])
        # wave_data.shape = (-1, 2)   # -1的意思就是没有指定,根据另一个维度的数量进行分割

        # 最后通过采样点数和取样频率计算出每个取样的时间
        time = np.arange(0, nframes) * (1.0 / framerate)

        plt.figure()
        # 左声道波形
        plt.subplot(2, 1, 1)
        # print(wave_data[:1000, 0].tolist())
        plt.plot(time, wave_data[:, 0])
        plt.xlabel("时间/s", fontsize=14)
        plt.ylabel("幅度", fontsize=14)
        plt.title("左声道", fontsize=14)
        plt.grid()  # 标尺

        plt.subplot(2, 1, 3)
        # 右声道波形
        plt.plot(time, wave_data[:, 1], c="g")
        plt.xlabel("时间/s", fontsize=14)
        plt.ylabel("幅度", fontsize=14)
        plt.title("右声道", fontsize=14)
        plt.tight_layout()  # 紧密布局
        # plt.show()

        aList = wave_data[:, 0].tolist()

        split = []
        round = 0

        iterm = []
        for i in range(len(aList)):
            if aList[i] < ignoreScore:
                round += 1
            if aList[i] >= 0.1:
                round = 0
            if aList[i] not in [0, 0.0] or round < 300 / 1000 * framerate:
                iterm.append(abs(aList[i]))
            if round > validTime / 1000 * framerate and aList[i] < ignoreScore:
                if len(iterm) >= (validTime / 1000 * framerate):
                    split.append(copy.deepcopy(iterm))
                round = 0
                iterm = []

        if len(iterm) >= (validTime / 1000 * framerate):
            split.append(iterm)

        # plt.show()
        return [sum(i) / len(i) for i in split if (sum(i) / len(i) > effectiveScore)]

    except:
        return []


if "__main__" == __name__:
    path = r"/left/test.wav"
    print(ProcessWav(path))