Reputation: 41
What would be an easy way to detect the speech portions of an audiostream in realtime:? I have sread about using MFCC's , FFTs and so on. But I couldnt find any clear explanation of how itis to be done.
Additionally, is there any PC program which can analyze wav files and detect the presence of human speech present in it?
Upvotes: 0
Views: 331
Reputation: 3844
This topic is so much complex...; there are no simple solutions or clear explanations; there are many existing techniques and algorithms. Modern techniques involve artificial intelligence and machine learning algorithms, not only trivial DSP algorithms like FFT or cepstral analysis. However, you can try Sphinx-4 which is written in Java so that it is naturally cross-platform.
http://cmusphinx.sourceforge.net/sphinx4/
I haven't tried that so I am not able to say anything about its usability, but be sure that the best algorithms are patented and really, really expensive...
Upvotes: 2