Software

ICMC FEATURES FOR UTTERANCE VERIFICATION AND TEXT-DEPENDENT ASV

Matlab implementation of Infinite impulse response constant Q Mel-frequency cepstral coefficients (ICMC) successfully used for utterance verification and text-dependent speaker verification. Additional supporting modules are included: voice activity detector, pre-emphasis filter and cepstral mean and variance normalisation. For further information consult the following paper:

Further optimisations of constant Q cepstral processing for integrated utterance verification and text-dependent speaker verification
Héctor Delgado, Massimiliano Todisco, Md Sahidullah, Achintya K. Sarkar, Nicholas Evans, Tomi Kinnunen, Zheng-Hua Tan, submitted to IEEE workshop on Spoken Language Technology (SLT) 2016, San Jose, Puerto Rico. [PDF]

Download the package from EURECOM Speech and Audio Processing Group site

CQCC-ARTE FEATURES FOR ASV

Matlab implementation of constant Q cepstral coefficients and articulation rate (ARTE) filter successfully used for automatic speaker verification. Additional supporting modules are added: voice activity detector, pre-emphasis filter and cepstral mean and variance normalisation. For further information consult the following paper:

Articulation rate filtering of CQCC features for automatic speaker verification
Massimiliano Todisco, Héctor Delgado and Nicholas Evans, in proc. INTERSPEECH 2016, San Francisco, USA. [PDF]

Download the package from EURECOM Speech and Audio Processing Group site

CQCC FEATURES FOR SPOOFED SPEECH DETECTION

Matlab implementation of constant Q cepstral coefficients successfully used for spoofed speech detection for speaker verification. For further information consult the following paper:

A new feature for automatic speaker verification anti-spoofing: Constant Q Cepstral Coefficients
Massimiliano Todisco, Héctor Delgado and Nicholas Evans, in proc. ODYSSEY 2016 – The Speaker and Language Recognition Workshop, Bilbao, Spain. [PDF]

Download the package from EURECOM Speech and Audio Processing Group site

BINARY KEY DIARIZER

Speaker diarization system based on the binary key speaker modeling, implemented in Matlab. For further information consult the following paper:

Fast single- and cross-show speaker diarization using binary key speaker modeling
Héctor Delgado, Xavier Anguera, Corinne Fredouille and Javier Serrano, in IEEE/ACM Transactions on Audio, Speech and Language Processing (TASLP). [PDF]

Download the package here