PROBLEM TO BE SOLVED: To solve the problem of a conventional technology for analyzing a vocal sound accompanied by vocal cord vibration based on AR-HMM, which used topology HMM having respective states connected in a ring shape because the vocal cord vibration is periodic: specifically, in the prior art having HMM topology connected in a ring shape, detailed analysis as to the quality of an alternative generated voice such as an esophageal speech not using any vocal cord and the quality deterioration of the voice is impossible.SOLUTION: HMM topology which is a sound source model is automatically generated bottom-up from a voice signal, thus enabling extraction of useful information capable of identifying the diseases of various vocal cords causing vocal troubles such as vocal cord polyps or larynx cancers, the individuality of a speaker, feeling included in a voice, the quality of alternative generated voice such as esophageal speech not using any vocal cord, and the quality deterioration of a voice.
