Browse Prior Art Database

Speaker Identification

IP.com Disclosure Number: IPCOM000092131D
Original Publication Date: 1968-Sep-01
Included in the Prior Art Database: 2005-Mar-05
Document File: 2 page(s) / 29K

Publishing Venue

IBM

Related People

Clapper, GL: AUTHOR

Abstract

This method of speaker identification comprises a number of steps for detecting a plurality of distinct speech features numbering from E1 to En. The occurrence of each of the plurality of distinct speech features over a time interval is summed. The plurality of distinct speech features is ranked according to the magnitude of the summation resulting from the summing. A binary code based on the rankings is produced. Such binary code is applied to an adapter catagorizer capable of learning and identifying individual speakers. The method of speech identification employed identifies features in the excitation function of speech. Speech analyzer 1 contains a harmonic locater which marks those points of the recurrent excitation function waveform having a local maximum or zero slope.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 90% of the total text.

Page 1 of 2

Speaker Identification

This method of speaker identification comprises a number of steps for detecting a plurality of distinct speech features numbering from E1 to En. The occurrence of each of the plurality of distinct speech features over a time interval is summed. The plurality of distinct speech features is ranked according to the magnitude of the summation resulting from the summing. A binary code based on the rankings is produced. Such binary code is applied to an adapter catagorizer capable of learning and identifying individual speakers. The method of speech identification employed identifies features in the excitation function of speech. Speech analyzer 1 contains a harmonic locater which marks those points of the recurrent excitation function waveform having a local maximum or zero slope. Scan ring 5 is started with the beginning of each glottal impulse and serves to identify the position of the local maximum as a function of the delay from the start of the excitation function. The summation, i.e., counting of these features is used to differentiate between different speakers. That is, the occurrence of each of the identification features over a period of time is counted by summation circuitry 2. The summation of the identification features is ranked and encoded in ranker and encoder 3 to produce a binary code for use in adaptive catagorizer 4. The latter is capable of learning a binary code which can be associated with a given speaker and afterwards ca...