Speaker verification utilising artificial neural networks and biometric functions derived from time encoded speech (TES) data

An initial investigation into the use of time encoded speech (TES) fixed size, fixed dimension A-matrices as the input layer of a simple fast artificial neural network (FANN), configured to identify the acoustic output of ten cooperative male imposters each articulating twenty versions of the same phase, `my name is Charles Westlake´, produced very favourable initial indications. The current investigation is an attempt to assess, to a first order, the suitability or otherwise, of a small (<10) set of fixed size biometric functions derived from TES `A-Matrices´ as the input layer for a simple artificial neural network configured to verify the identity of a given speaker from nine `imposters´. The ability of these TES biometric/FANN combinations to converge with training data is investigated and their performance against different interrogation data is evaluated