Python实现简单的语音识别系统_Python

最近认识了一个做Python 语音识别的朋友，聊天时候说到，未来五到十年，Python人工智能会在国内掀起一股狂潮，对各种应用的冲击，不下于淘宝对实体经济的冲击。在本地（江苏某三线城市）做这一行，短期可能显不出效果，但从长远来看，绝对是一个高明的选择。朋友老家山东的，毕业来这里创业，也是十分有想法啊。

将AI课上学习的知识进行简单的整理,可以识别简单的0-9的单个语音。基本方法就是利用库函数提取mfcc,然后计算误差矩阵,再利用动态规划计算累积矩阵。并且限制了匹配路径的范围。具体的技术网上很多,不再细谈。

现有缺点就是输入的语音长度都是1s,如果不固定长度则识别效果变差。改进思路是提取有效语音部分。但是该部分尚未完全做好,只写了一个原形函数,尚未完善。

Python实现简单的语音识别系统

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

									import wave

									import numpy as np

									import matplotlib.pyplot as plt

									from python_speech_features import mfcc

									from math import cos,sin,sqrt,pi

									def read_file(file_name):

									  with wave.open(file_name,'r') as file:

									    params = file.getparams()

									    _, _, framerate, nframes = params[:4] 

									    str_data = file.readframes(nframes)

									    wave_data = np.fromstring(str_data, dtype = np.short)

									    time = np.arange(0, nframes) * (1.0/framerate)

									    return wave_data, time 

									  return index1,index2

									def find_point(data):

									  count1,count2 = 0,0

									  for index,val in enumerate(data):

									    if count1 <40:

									      count1 = count1+1 if abs(val)>0.15 else 0

									      index1 = index

									    if count1==40 and count2 <5:

									      count2 = count2+1 if abs(val)<0.001 else 0

									      index2 = index

									    if count2==5:break

									  return index1,index2

									def select_valid(data):

									  start,end = find_point(normalized(data))

									  print(start,end)

									  return data[start:end]

									def normalized(a):

									  maximum = max(a)

									  minimum = min(a)

									  return a/maximum

									def compute_mfcc_coff(file_prefix = ''):

									  mfcc_feats = []

									  s = range(10)

									  I = [0,3,4,8]

									  II = [5,7,9]

									  Input = {'':s,'I':I,'II':II,'B':s}

									  for index,file_name in enumerate(file_prefix+'{0}.wav'.format(i) for i in Input[file_prefix]):

									    data,time = read_file(file_name)

									    #data = select_valid(data)

									    #if file_prefix=='II':data = select_valid(data)

									    mfcc_feat = mfcc(data,48000)[:75]

									    mfcc_feats.append(mfcc_feat)

									  t = np.array(mfcc_feats)

									  return np.array(mfcc_feats)

									def create_dist():

									  for i,m_i in enumerate(mfcc_coff_input):#get the mfcc of input

									    for j,m_j in enumerate(mfcc_coff):#get the mfcc of dataset

									      #build the distortion matrix bwtween i wav and j wav

									      N = len(mfcc_coff[0])

									      distortion_mat = np.array([[0]*len(m_i) for i in range(N)],dtype = np.double)

									      for k1,mfcc1 in enumerate(m_i):

									        for k2,mfcc2 in enumerate(m_j):

									          distortion_mat[k1][k2] = sqrt(sum((mfcc1[1:]-mfcc2[1:])**2))

									      yield i,j,distortion_mat

									def create_Dist():

									  for _i,_j,dist in create_dist():

									    N = len(dist)

									    Dist = np.array([[0]*N for i in range(N)],dtype = np.double)

									    Dist[0][0] = dist[0][0]

									    for i in range(N):

									      for j in range(N):

									        if i|j ==0:continue

									        pos = [(i-1,j),(i,j-1),(i-1,j-1)]

									        Dist[i][j] = dist[i][j] + min(Dist[k1][k2] for k1,k2 in pos if k1>-1 and k2>-1)

									    #if _i==0 and _j==1 :print(_i,_j,'\n',Dist,len(Dist[0]),len(Dist[1]))

									    yield _i,_j,Dist

									def search_path(n):

									  comparison = np.array([[0]*10 for i in range(n)],dtype = np.double)

									  for _i,_j,Dist in create_Dist():

									    N = len(Dist)

									    cut_off = 5

									    row = [(d,N-1,j) for j,d in enumerate(Dist[N-1]) if abs(N-1-j)<=cut_off]

									    col = [(d,i,N-1) for i,d in enumerate(Dist[:,N-1]) if abs(N-1-i)<=cut_off]

									    min_d,min_i,min_j = min(row+col )

									    comparison[_i][_j] = min_d

									    optimal_path_x,optimal_path_y = [min_i],[min_j]

									    while min_i and min_j:

									      optimal_path_x.append(min_i)

									      optimal_path_y.append(min_j)

									      pos = [(min_i-1,min_j),(min_i,min_j-1),(min_i-1,min_j-1)]

									      #try:

									      min_d,min_i,min_j = min(((Dist[int(k1)][int(k2)],k1,k2) for k1,k2 in pos\

									      if abs(k1-k2)<=cut_off))

									    if _i==_j and _i==4:

									      plt.scatter(optimal_path_x[::-1],optimal_path_y[::-1],color = 'red')

									      plt.show()

									  return comparison

									mfcc_coff_input = []

									mfcc_coff = []

									def match(pre):

									  global mfcc_coff_input

									  global mfcc_coff

									  mfcc_coff_input = compute_mfcc_coff(pre)

									  compare = np.array([[0]*10 for i in range(len(mfcc_coff_input))],dtype = np.double)

									  for prefix in ['','B']:

									    mfcc_coff = compute_mfcc_coff(prefix)

									    compare += search_path(len(mfcc_coff_input))

									  for l in compare:

									    print([int(x) for x in l])

									    print(min(((val,index)for index,val in enumerate(l)))[1])

									data,time = read_file('8.wav')

									match('I')

									match('II')