语音特征提取（语谱图Spectrogram，Fbank, MFCC, 及其delta-一阶差分）——python代码

导入相关包

import os
import wavio
import numpy as np
import math
from matplotlib import pyplot as plt
from scipy.fftpack import dct
from python_speech_features import mfcc, delta, logfbank

读取语音数据及主函数

for wav in wavs:
	wav_dir = os.path.join(data_dir, wav)
	wav_data = wavio.read(wav_dir)
	data = wav_data.data
	sample_rate = wav_data.rate	#16k
	sampwidth = wav_data.sampwidth
	
	#normalization
	norm_data = data/max(abs(data))
	
	#frames
	frames = frames_crop(norm_data,sample_rate)
	
	#add window
	win = 160
	windows = choose_windows(name = "Hamming",N = win)
	
	#parameters
	N = 2048 #NFFT
	M = 40 #filters number
	num_ceps = 24
	
	#fft
	spe_freqs = np.zeros((frames.shape[0],int(N/2)))	#spectrogram
	fbank_feature = np.zeros((frames.shape[0],M))	#Fbank
	fbank_feature_2 = np.zeros((frames.shape[0],M))	#Fbank second version
	
	mfcc_dct = np.zeros((frames.shape[0],num_ceps))	#dct
	
	
	for i in range(frames.shape[0]):
		frames_fft = np.fft.fft(windows * frames[i],N)
		spe_freqs[i][:] = log_data(np.abs(frames_fft[:int(N/2)]))
		
		filter_banks = mel_filters(sample_rate = sample_rate, NFFT = N, pow_frames = np.abs(frames_fft[:int(N/2