Automatic speaker recognition based upon various distances of LSP frequencies

A way to identify people by voice is discussed. A speaker's spectral information represented by line spectrum pair (LSP) frequencies is used to describe characteristics of the speaker's utterance and the VQ (vector quantization) method is used to model the spectral distribution of each speaker. Some easily computed distances (Euclidean distance, weighted distance by F-ratio, and hard-limited distance) are used to measure the discrimination among speakers.<
>