Extracting human voice from wav file using java

I'm working on a project where I have to use Java from audio Extract human voice from wav file

Audio WAV files may have 3 to 4 sounds, such as dogs, cats, music and humans I will have to recognize the human voice and then from the audio Delete this section from the wav file

I'm using FFT Java and complex java.

Now I write an audiofilereader class that reads audio from the hard drive Wav file, and then convert it to a byte array Then use the FFT mentioned above Java and complex Java application FFT FFT (bytesarray), which gives the return value of complex array;

The question now is how to extract human voice byte patterns from the returned complex array... Does anyone know how I can achieve this?

Editor: let's assume a very simple audio Wav file For example, the cat's voice is then silent, the human voice is then silent, the dog's voice is then silent and so on

Solution

I think the standard method to deal with this problem is to convert the input signal into cepstrum or Mel cepstrum representation, and then input the coefficients of feature space into the classifier Many research papers have discussed the methods to solve these problems based on this basic method, such as:

http://www.ics.forth.gr/netlab/data/J17.pdf

One possible shortcut you may try is to decode the input signal through a low bit rate vocoder such as AMBE, and then compare the quality of the original signal with the encoded / decoded signal These vocoders are designed to compress human speech with high quality at the cost of not fully representing nonverbal sound