2
  • There are multiple speakers.
  • Each speaker generates multiple MFCC.
  • If I'm getting two different MFCC, can we know if those MFCC came from 2 different speakers or same speaker ?

How can we compare them and check this ?

Gilles
  • 3,386
  • 3
  • 21
  • 28
Boom
  • 121
  • 1
  • A naive approach: https://dsp.stackexchange.com/a/40830/8202 https://dsp.stackexchange.com/a/29167/8202 – jojeck Jan 12 '21 at 12:28

1 Answers1

-1
  • You can find the distance between the MFCC features for different speakers. The distance should be zero for the same speaker and larger with different speakers.
  • You can use different types of distance functions or even similarity functions used generally to compare between probability density functions.
  • Also, corresponding to each speaker you should generate a single matrix of MFCC coefficients which is composed of multiple coefficients for each frame. So, the output MFCC should be of dimension: (number of frames, number of coefficients).
  • Every speaker's MFCC dimension should be the same to compare them.
Aaquila
  • 32
  • 4
  • I look on 2 different MFCC matrix of same speaker, and the values are different. so we need more sophisticated way to compare – Boom Dec 13 '20 at 13:05
  • Was both of them obtained from the same audio recording? – Aaquila Dec 14 '20 at 18:47
  • Usually, these features change with the words and the way u speak so essentially unless it is the same recording it may not be same. – Aaquila Dec 14 '20 at 18:48