top of page

Different Colours for different Speakers

Speaker Diarization

The speaker diarization uses unsupervised algorithm to detect number of speakers in the audio and then uses 2, 3, 4 speaker diarization algorithm to separate the speaker audios. There is also a provision for manual diarization if the audio contains more than or equal to 5 speakers.

salient features of the system:

  1. Fully unsupervised system, detects number of speakers and then diarization is done

  2. The system is Language independent, accent independent

  3. Finds speaker embeddings for healthy speech segments after separating ring, music and silence segments.

  4. Works well for 2, 3 and 4 speakers

  5. Can be used to find the rate of speech for agent or client.

  6. Automatically detects the gender for segregated speakers.

bottom of page