Python实现简单的语音识别系统

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。

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')

总结

Python用户推荐系统曼哈顿算法实现完整代码

Python编程使用tkinter模块实现计算器软件完整代码示例

如有不足之处，欢迎留言指出。感谢朋友们对本站的支持！

来源：http://blog.csdn.net/pp634077956/article/details/52916699