Geek Answers Handbook

Audio connection (audio fingerprint)

I am writing an Android application that allows a user to record their voice through a microphone and store it in storage and associate it with specific content (e.g. Contact). Later, the user will call this voice again, and the application should compare it with the stored audio files and find the one that matches the voice.

I searched a lot and found several libraries that do this on the Internet, for example EchoPrint , which generates a fingerprint from recorded audio and sends it to the openource server and returns the result. But I need to do this offline.

Does anyone know such a library?

+4
source share
2 answers

If you are trying to compare an old user record with a new call, as it happens, sound fingerprint solutions like Dejavu in Python on the server or Echoprint in C ++ will not help. They are designed to recognize and search on recorded sound segments plus noise. They cannot deal with variables in the human voice. See explanation here .

If in this case we are talking about speaker recognition , it is much more complicated and requires quite a lot of machine learning. It would be difficult to do this for a large body of users (especially offline on the phone), but for the determination between two users this can be feasible.

+6

More articles:


All Articles