Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

questions bout chinese speech recognition in LPC & ANN

Status
Not open for further replies.

qdipanda

Newbie level 1
Joined
Sep 8, 2009
Messages
1
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,294
ann lpc voice recognition

i hav few questions and problems need help
i'm currently doin the chinese speech recognition using LPC ans ANN..

the way i doin nw is lik below
Recognition process
1-record sound fr a user
2-perform feature extraction
-sound normalization
-noise reduction
-VAD, voice activity detection
3-LPC coeff generation, coeff = lpc(sound,n-1); //n=10
4-coeff use for recognition , [a,Pf,Af]=sim(net,coeff);


Training process
1-gather few set of sound fr few ppl, eg, 一...八(1,2,3....8 ) gather 4 set as target
2-perform feature extraction
-sound normalization
-noise reduction
-VAD, voice activity detection
3-LPC coeff generation, coeff = lpc(sound,n-1); //n=10
4-perform training process useing ANN

net=newff(minmax(inputs),[21 1],{'tansig','purelin'},'traingd');
net=train(net,inputs,targets);


my problem nw is the system not recognize wat i say....

So follow is d question i need to ask...
1) is it d method at above i doin is correct o wrong??
2) if i do d correct way, d network training process is it do lik dat??
3) after the NN training process...,the regression, performance and training state of d neural network should look lik hw??
 

chinese speech recognition

It is correct, but I suggest you to use PLP or MF-PLP for feature
extraction.
MFCC is also popular.

HTK (https://htk.eng.cam.ac.uk/) has MFCC, MF-PLP, LPC and other
techniques.
Also consider to use HMMs (Hidden Markov Models). HMMs are very
popular for
speech recognition because training process is robust.

Luis Uebel
https://www.asrlabs.com
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top