SpeakerVerify - Education Tool

SpeakerVerify is a real-time Automatic Speaker Verification and Recognition education tool. It has been created to demonstrate to students studying Digital Signal Processing (DSP) how a basic feature based Automatic Speaker Verification (ASV) systems works.

SpeakerVerify decomposes the speaker verification problem into signal processing blocks, allowing the user to change individual parameters and directly identify its effects on the verification decision. SpeakerVerify trains on one speaker and tests on one speaker.

##### Technical Information #######

Signal Processing settings:

- Pre-Emphasis

- Frame Length

- Frame Shift

- Silence Removal

- Hamming analysis window

- FFT size

Feature Extraction Settings:

- LPC

- LPCC

- RC

- LAR

- ARC

- MFCCs

- Number of filter banks

Classification / Pattern Matching settings:

- Euclidean Distance (Selectable threshold)

- Basic single mixture GMM (Trained offline)

This is a beta version, if you have any problems please contact the developer with the "email developer" button on this page.