TARGET SPECIFIC DATA FILTER TO SPEED PROCESSING

• US 20090043579A1
• Filed: 04/02/2008
• Published: 02/12/2009
• Est. Priority Date: 07/07/2006
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Verscheure, Olivier, Huerta, Juan M., Ramaswamy, Ganesh N., Chaudhari, Upendra V.
• Granted Patent: US 7,831,424 B2
• Time in Patent Office: Days
• 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

A method is presented which reduces data flow and thereby increases processing capacity while preserving a high level of accuracy in a distributed speech processing environment for speaker detection. The method and system of the present invention includes filtering out data based on a target speaker specific subset of labels using data filters. The method preserves accuracy and passes only a fraction of the data by optimizing target specific performance measures. Therefore, a high level of speaker recognition accuracy is maintained while utilizing existing processing capabilities.

Claims

1. A method for reducing the amount of speech data for detecting desired speakers, comprising:

receiving a plurality of speech data from a plurality of speakers;

labeling the speech data;

providing a speaker dependent filter which employs a label priority ranking of the labeled speech data for each of the speakers, and the ranking is generated by prioritizing the labeled speech data according to a performance criterion for each of the plurality of speakers; and

analyzing the labeled speech data according to the priority rank of the label of the speech data for a specific speaker, wherein specified lower priority labeled speech data is dropped or filtered out such that less speech data is analyzed.

2. The method of claim 1, wherein the speakers are divided into groups of speakers and the filters are generated and used for each group of speakers.

3. The method of claim 2, wherein the labels for the groups of speakers are priority ranked.

4. The method of claim 1, wherein the data is pre-labeled.

5. A speaker recognition system using a computer, which comprises:

a plurality of speech data from a plurality of speakers stored on a computer;

a labeler element for analyzing the speech data and labeling the data,

a speech data filter including a filter element for arranging the labels in a priority rank and the speech data filter designed according to a performance criterion for each of the plurality of speakers; and

a recognition element for analyzing the labeled data in accordance with the priority rank as determined by the speech data filter.

6. The system of claim 5, wherein the labels are phonetic.