Our voice biometrics systems allow automatic speaker recognition to be performed in a flexible way, using an ‘open-box’ architecture and providing a choice of several feature extraction, speaker modelling, and speaker comparison approaches, and supporting the training and adaptation of models with user data at multiple points of the speaker recognition pipeline. Our systems offer forensic speaker comparison, speaker identification and verification using GMM, i-vectors, and state-of-the-art deep neural network-based x-vectors.
Our performance metrics software can be used to interpret the scores returned from ASR and can also be used to estimate a likelihood ratio (LR) under same-speaker and different-speaker hypotheses, enabling the forensic practitioner to make objective estimates of the strength of the evidence in a speaker comparison case.
Windows | API
Highly reliable voice biometrics software to perform fast, accurate speaker comparisons using Deep Neural Network based x-vectors. Adaptable to your applications. Powerful visualisation and evidence handling.
Easily evaluate the performance of your speaker biometric systems or algorithms to obtain error rates and likelihood ratios from results to reports in seconds.
Powerful, web-based, networked, high speed voice biometrics system, suited for large 1:N speaker comparisons against a centralised database of suspect voice recordings.
On-device, mobile, handheld voice biometrics: Automatic speaker recognition and profiling systems for reliably determining speaker identity in the field. Comparisons performed fully on-device, independent of any connectivity.