top of page

Forensic Speaker identification

The speaker verification and identification system works for text dependent and text independent scenario. It is language independent system. If the text in the two files used for verification is same, system extracts features like formants for word pairs extracted from text and verifies a speaker. For text independent case, the system directly compares the audios by comparing the embedding vectors.

Salient features of the system:

  1. Uses word separation tool for segregating words in common text for suspect and control audio.

  2. Extracts and compares the formants in the two word pairs automatically and generates the pdf report for speaker verification.

  3.  For automatic speaker verification with text independent scenario, it extracts features like embedding and directly gives matching score for speaker verification.

If the parameters for number of culprits are already stored in the system, it can be used for speaker identification as well.

bottom of page