【yyddjtc】MFCC算法--python实现

import numpy as np
import librosa
import matplotlib.pyplot as plt
import scipy.io.wavfile
from scipy.fftpack import dct
data,sr = librosa.load('../recordings/1_george_8.wav')

def mfcc(data,sr):
        #预加重
        l = np.append(data[0],data[1:]-0.97*data[:-1])
        #分帧
        frame_size = 0.025 #25ms 为一帧
        frame_stride = 0.01 # 10ms 滑动（15ms的交叠部分）
        frame_length  = int(round(frame_size * sr)) #帧的大小
        frame_step = int(round(frame_stride * sr))  #帧的滑动
        num_frames = int(np.ceil(float(np.abs(len(data) - frame_length)) / frame_step)) #有多少帧
        indices = np.tile(np.arange(0, frame_length), (num_frames, 1)) + np.tile(
                np.arange(0, num_frames * frame_step, frame_step), (frame_length, 1)).T
        #indices 中的每一个元素为一帧,一帧有550个值，滑动220
        frames = l[np.mat(indices).astype(np.int32, copy=False)]
        #为frames添加窗函数
        frames *= np.hamming(frame_length)
        #傅里叶变换
        NFFT = 512
        mag_frames = np.absolute(np.fft.rfft(frames, NFFT))
        #输出热力谱
        pow_frames = ((1.0 / NFFT) * ((mag_frames) ** 2))
        #转换mel频谱
        h = (2595 * np.log10(1 + (sr / 2) / 700)) #把最高频率转化为mel频率
        M = 40
        mel_points = np.linspace(0, h, M+2)  # 以mel频率为准划分坐标
        hz_points = (700 * (10 ** (mel_points / 2595) - 1))  # 再把坐标的点转换为正常频率
        bin = np.floor((NFFT + 1) * hz_points / sr) #hz_point的索引
        fbank = np.zeros((M, int(np.floor(NFFT / 2 + 1)))) #一个40*257的容器
        #加三角窗
        for m in range(1, M + 1):
                f_m_minus = int(bin[m - 1])  # left
                f_m = int(bin[m])  # center
                f_m_plus = int(bin[m + 1])  # right
                for k in range(f_m_minus, f_m):
                        fbank[m - 1, k] = (k - bin[m - 1]) / (bin[m] - bin[m - 1])
                for k in range(f_m, f_m_plus):
                        fbank[m - 1, k] = (bin[m + 1] - k) / (bin[m + 1] - bin[m])
        #对热力谱施加mel变换和三角窗
        filter_banks = np.dot(pow_frames, fbank.T)
        #数值稳定性
        filter_banks = np.where(filter_banks == 0, np.finfo(float).eps, filter_banks)
        #以分贝为单位
        filter_banks = 20 * np.log10(filter_banks)
        #DCT变换
        mfcc = dct(filter_banks, type=2, axis=1, norm='ortho')[:, 1 : 13]
        #减去均值
        mfcc -= (np.mean(mfcc, axis=0) + 1e-8)
        # 画热力图
        plt.imshow(np.flipud(mfcc.T), cmap=plt.cm.jet, aspect=0.2, extent=[0, mfcc.shape[1], 0, mfcc.shape[0]])
        plt.show()
        return mfcc

print(mfcc(data,sr))

输入为音频时间序列和采样率，输出为mfcc可用图像展示出来。

整理自：

https://thnum.blog.csdn.net/article/details/80597495https://thnum.blog.csdn.net/article/details/80597495