VOICE DEGREE OF SIMILARITY DETECTING METHOD AND VOICE RECOGNITION METHOD USING THE DETECTED VALUE, VIBRATING WAVE DEGREE OF SIMILARITY DETECTING METHOD AND MACHINE ABNORMAL DISCRIMINATING METHOD USING THE DETECTED VALUE, IMAGE DEGREE OF SIMILARITY DETECTING METHOD AND IMAGE RECOGNITION METHOD USING THE DETECTED VALUE, SOLID BODY DEGREE OF SIMILARITY DETECTING METHOD AND SOLID BODY RECOGNITION METHOD USING THE DETECTED VALUE AND ANIMATION IMAGE DEGREE OF SIMILARITY DETECTING METHOD AND ANIMATION IMAGE RECOGNITION METHOD USING THE DETECTED VALUE

摘要

PROBLEM TO BE SOLVED: To provide a method by which voice similarity or the like is accurately detected. SOLUTION: Standard and input pattern matrices having voice features as components are generated (S1 to S2). Then, a reference shape having a different dispersion value for every specified component of the pattern matrices is generated. Then, positive and negative reference pattern vectors having reference shape values as components are generated. Then, shape variations between the standard and input pattern matrices are replaced by the shape variations of the positive and negative reference pattern vectors while successively matching specified components (centers of the reference shapes) to J1=1 to m1, J2=1 to m2 component positions of the standard pattern matrix. The amount of change in the degree of sharpness of each reference pattern vector is digitized an image into a shape change amount Dj1j2 (S3) and a shape distance value between the pattern matrices is computed from Dj1J2 (S4).