【语音识别】基于matlab GUI BP神经网络0到10数字语音识别【含Matlab源码 672期】

一、BP神经网络语音识别简介

1 对语音的WAV文件和LAB文件进行处理，产生十个文件，每个文件对应于一个数字，存贮着该数字的波形文件。（shujuzhengli.m）
2 分别利用上面十个文件训练出十个HMM模板，具体方法是：首先将语音的波形文件分帧，以128个点为一帧，帧为64，每一帧通过mfcc.m计算出13个系数，随着波形的长度不同，一个语音文件可以计算得到13N个系数，截取1315的矩阵（mfcc系数）用作训练数据。一般一个HMM模板用20组mfcc系数训练，得到初始状态分布、状态转移矩阵、高斯正态分布的均值和方差以及混合矩阵，这就是该语音的特征，存贮下来，识别的时候使用。（trainmfcc.m）
3 识别过程
识别的前面部分与训练相似，都是要计算得到mfcc系数，不同在于，识别时，将计算得到的mfcc 参数分别代入训练得到的HMM模板求出概率，比较出最大概率者，则该模板对应的数字就是识别的数字。(shibiesb.m)
4 用大量语音文件做测试，结果正确率为90 以上。

二、部分源代码

clc
cleark = 11;data_matrix=[];%保存训练数据data_matrix2=[];%保存测试数据
%载入训练数据，并提取特征
T_train=[];
T_test=[];
for i=0:9for j=10:50
s = sprintf('09517/%i%i.wav',i,j);%把格式化的数据写入某个字符串中
[s1 fs1] = audioread(s);%读取
v = mfcc(s1, fs1);%提取特征参数
a= vqlbg(v, k); %量化
data_matrix=[data_matrix,a(:)];
T_train=[T_train,i+1];end
end%载入测试数据，并提取特征
for i=0:9for j=51:99
s = sprintf('09517/%i%i.wav',i,j);
[s1 fs1] = audioread(s);
v = mfcc(s1, fs1);
a= vqlbg(v, k);
data_matrix2=[data_matrix2,a(:)];
T_test=[T_test,i+1];end
end
%样本的标签% P_train=mapminmax(data_matrix,0,1)';%归一
[P_train,settings] = mapminmax(data_matrix,0,1);
[P_test]=mapminmax(data_matrix2,'aply',settings);%归一
Tn_train=BP(T_train');
Tn_test=BP(T_test');
P_train=P_train;
P_test=P_test;
net=newff(minmax(P_train),[200,10],{'tansig' 'tansig'} ,'traingda');%建立一个神经网络框架
net.trainParam.show=500;
%训练网络
net.trainParam.lr=0.5;
net.trainParam.epochs=25000;      %训练次数取5000
net.trainParam.goal=0.01;        %误差门限取0.001
net=train(net,P_train,Tn_train); %训练神经网络
YY=sim(net,P_train);
[maxi,ypred]=max(YY);
maxi=maxi';
ypred=ypred';
CC=ypred-T_train';
n=length(find(CC==0));
Accuracytrain=n/size(P_train,2)%算识别的标签和真实的标签不一样的个数,从而计算出正确率YY=sim(net,P_test);
[maxi,ypred]=max(YY);
maxi=maxi';
ypred=ypred';
CC=ypred-T_test';
n=length(find(CC==0));
Accuracytest=n/size(P_test,2)
function varargout = yuyinshibie(varargin)
% YUYINSHIBIE M-file for yuyinshibie.fig
%      YUYINSHIBIE, by itself, creates a new YUYINSHIBIE or raises the existing
%      singleton*.
%
%      H = YUYINSHIBIE returns the handle to a new YUYINSHIBIE or the handle to
%      the existing singleton*.
%
%      YUYINSHIBIE('CALLBACK',hObject,eventData,handles,...) calls the local
%      function named CALLBACK in YUYINSHIBIE.M with the given input arguments.
%
%      YUYINSHIBIE('Property','Value',...) creates a new YUYINSHIBIE or raises the
%      existing singleton*.  Starting from the left, property value pairs are
%      applied to the GUI before yuyinshibie_OpeningFcn gets called.  An
%      unrecognized property name or invalid value makes property application
%      stop.  All inputs are passed to yuyinshibie_OpeningFcn via varargin.
%
%      *See GUI Options on GUIDE's Tools menu.  Choose "GUI allows only one
%      instance to run (singleton)".
%
% See also: GUIDE, GUIDATA, GUIHANDLES% Edit the above text to modify the response to help yuyinshibie% Last Modified by GUIDE v2.5 08-Aug-2017 22:59:02% Begin initialization code - DO NOT EDIT
gui_Singleton = 1;
gui_State = struct('gui_Name',       mfilename, ...'gui_Singleton',  gui_Singleton, ...'gui_OpeningFcn', @yuyinshibie_OpeningFcn, ...'gui_OutputFcn',  @yuyinshibie_OutputFcn, ...'gui_LayoutFcn',  [] , ...'gui_Callback',   []);
if nargin && ischar(varargin{1})gui_State.gui_Callback = str2func(varargin{1});
endif nargout[varargout{1:nargout}] = gui_mainfcn(gui_State, varargin{:});
elsegui_mainfcn(gui_State, varargin{:});
end
% End initialization code - DO NOT EDIT% --- Executes just before yuyinshibie is made visible.
function yuyinshibie_OpeningFcn(hObject, eventdata, handles, varargin)
% This function has no output args, see OutputFcn.
% hObject    handle to figure
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
% varargin   command line arguments to yuyinshibie (see VARARGIN)% Choose default command line output for yuyinshibie
handles.output = hObject;% Update handles structure
guidata(hObject, handles);% UIWAIT makes yuyinshibie wait for user response (see UIRESUME)
% uiwait(handles.figure1);% --- Outputs from this function are returned to the command line.
function varargout = yuyinshibie_OutputFcn(hObject, eventdata, handles)
% varargout  cell array for returning output args (see VARARGOUT);
% hObject    handle to figure
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)% Get default command line output from handles structure
varargout{1} = handles.output;% --- Executes on button press in pushbutton1.
function pushbutton1_Callback(hObject, eventdata, handles)
global str;
global s1;
global fs1;
[filename,pathname]=...uigetfile({'*.wav';'*.bmp';'*.gif'},'choose');
str=[pathname filename];
[s1 fs1] = wavread(str);%读取
% a=imread(str);
% 同学在这里写上进度条的代码 等待对话框
h=waitbar(0,'Pleast waiting, reading...');
axes(handles.axes1);
plot(s1)
waitbar(1,h,'finish');
pause(0.05);
delete(h);
function [x1,x2,FrameInc,amp,zcr] = vad(x) %幅度归一化到[-1,1]
x = double(x);
x = x / max(abs(x)); %常数设置
FrameLen = 240;
FrameInc = 80; amp1 = 10;
amp2 = 2;
zcr1 = 10;
zcr2 = 5; maxsilence = 8;  % 6*10ms  = 30ms
minlen  = 15;    % 15*10ms = 150ms
status  = 0;
count   = 0;
silence = 0; %计算过零率
tmp1  = enframe(x(1:end-1), FrameLen, FrameInc);
tmp2  = enframe(x(2:end)  , FrameLen, FrameInc);
signs = (tmp1.*tmp2)<0;
diffs = (tmp1 -tmp2)>0.02;
zcr   = sum(signs.*diffs, 2); %计算短时能量
amp = sum(abs(enframe(filter([1 -0.9375], 1, x), FrameLen, FrameInc)), 2); %调整能量门限
amp1 = min(amp1, max(amp)/4);
amp2 = min(amp2, max(amp)/8); %开始端点检测
x1 = 0;
x2 = 0;
for n=1:length(zcr) goto = 0; switch status case {0,1}                   % 0 = 静音, 1 = 可能开始 if amp(n) > amp1          % 确信进入语音段 x1 = max(n-count-1,1); status  = 2; silence = 0; count   = count + 1; elseif amp(n) > amp2 | ... % 可能处于语音段 zcr(n) > zcr2 status = 1; count  = count + 1; else                       % 静音状态 status  = 0; count   = 0; end case 2,                       % 2 = 语音段 if amp(n) > amp2 | ...     % 保持在语音段 zcr(n) > zcr2 count = count + 1; else                       % 语音将结束 silence = silence+1; if silence < maxsilence % 静音还不够长，尚未结束 count  = count + 1; elseif count < minlen   % 语音长度太短，认为是噪声 status  = 0; silence = 0; count   = 0; else                    % 语音结束 status  = 3; end end case 3, break; end
end

三、运行结果

四、matlab版本及参考文献

1 matlab版本
2014a

2 参考文献
[1]韩纪庆,张磊,郑铁然.语音信号处理（第3版）[M].清华大学出版社，2019.
[2]柳若边.深度学习:语音识别技术实践[M].清华大学出版社，2019.