Voice verification system based on word pronunciation

• US 4,100,370 A
• Filed: 12/13/1976
• Issued: 07/11/1978
• Est. Priority Date: 12/15/1975
• Status: Expired due to Term

• Current Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Inventors: Suzuki, Matsumi, Morino, Tetsuro, Kitamoto, Saburo
• Primary Examiner(s): Brown, Thomas W.
• Assistant Examiner(s): Kemeny, E. S.
• Application Number:
  US05/750,323
• Time in Patent Office: 575 Days
• Field of Search: 179/1 SB, 179/1 SD, 340/146.3 FT, 340/146.3 T
• US Class Current: 704/246
• CPC Class Codes: G10L 17/02 Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction

Abstract

This system verifies the voice of a specific speaker (A) through verification of vocal sounds which form a previously-recorded recognition phrase spoken by A and others. Parallel filters derive a spectral characteristic parameter x_i which contains weighting factors α_i, β_i, and which represents a corresponding vocal sound x_i. Improved specificity over other speakers is attained by a Characteristic Feature Extractor which varies α_i, β_i, through p different values, thus deriving a set of p parameters (x_ik) for each vocal sound x_i, thus allowing emphasis of the specific vocal sound x_i as spoken by a specific speaker, over that of competing sounds and speakers.

Claims

1. A word pronunciation voice verifying system, comprising:

a voice level standardizing unit for standardizing the input voice level of a speaker whose voice is to be verified, a frequency analyzer unit for analyzing the standardized voice signal and including a plurality of parallel channels each having a different band pass frequency, an output level detector unit for detecting the output level of each of the plurality of channels, a characteristic feature extracting unit for extracting a plurality of characteristic parameters for one vocal sound from the detected output, a selective gate unit for selecting the parameter to be transmitted, based on a parameter selective signal, out of a plurality of characteristic parameters consisting of binary time series patterns for each of the input voices extracted by the characteristic feature extracting unit, a memory unit for storing the characteristic parameter pattern which has passed through the selective gate unit, a reference parameter memory unit for storing the binary time series pattern of the reference parameter, resemblance calculator units for calculating the resemblance between the parameters of the patterns stored in the memory unit and the reference parameter memory unit, respectively, and for calculating the resemblance between the input pattern which has passed through the selective gate and the time series pattern stored in the reference parameter memory unit, a suitable parameter discriminating unit for selecting a suitable parameter corresponding to a maximum resemblance based on the results of the calculation of the resemblance between the contents stored in the memory unit and the reference parameter memory unit by the resemblance calculator units, and for sending out a selective signal relative to the suitable parameter to the selective gate unit and the memory unit, and a discriminating and output unit for discriminating the vocabulary with regard to the pattern of the reference parameter corresponding to the maximum resemblance derived from the results of the calculation of the resemblance between the input pattern and the reference parameter by the resemblance calculator units, and for sending out the result of the discrimination as an output.