PiD, a person identification system using voice and face
USA Patent 5,412,738


The strategy used by the described system is based on Vector Quantization ( see the paper Person Recognition Using Multiple Cues,R. Brunelli and D. Falavigna). The acoustic vectors are compared with the codebooks of the reference speakers. The comparisons consists in the evaluation of a distance between input vectors and codebooks. The obtained distances (one for each reference user) are sent to the integration module.

Back to PiD homepage