how to train the MFCC data for speech recognition ?

1 次查看（过去 30 天）

Priya Singh 2017-2-9

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/324248-how-to-train-the-mfcc-data-for-speech-recognition

hello everyone im doing a speech recognition problem for that i have 5 samples and i have also calculated the mfcc for each of the samples. Im getting a 13*64 size mfcc for each sample, now to train it in ANN i thought of optimize the mfcc by using the PSO (particle swarm optimization) but im confused at should i make the mfcc (13*1) size or (1*64). i mean how i should optimise the mfcc ?? 64 is the no. of frames and 13 is the cepstral coeff i.e. for each frame im getting 13 coeff. therefore should i optimze it in frame wise i.e. for each frame instead of 13 only 1 coeff (total of 64 final coeff) or for all 64 frames the first mfcc will be 1 instead of 64 (finally will get a total of 13 coeff) ............please help