Prefix methods for diarization in streaming mode

Patent metadata

Field	Value
Publication number	US-2018158451-A1
Application number	US-201715827934-A
Country	US
Kind code	A1
Filing date	Nov 30, 2017
Priority date	Dec 1, 2016
Publication date	Jun 7, 2018
Grant date	—

Abstract

A diarization embodiment may include a system that clusters data up to a current point in time and consolidates it with the past decisions, and then returns the result that minimizes the difference with past decisions. The consolidation may be achieved by performing a permutation of the different possible labels and comparing the distance. For speaker diarization, a distance may be determined based on a minimum edit or hamming distance. The distance may alternatively be a measure other than the minimum edit or hamming distance. The clustering may have a finite time window over which the analysis is performed.

First claim

1 . An apparatus comprising:
a memory storing data; and a processor in communication with the memory, the processor configured to cluster data to a point in time and consolidates the data with a past decision value, and to output a result that minimizes a difference with the past decision value.

2 . The apparatus of claim 1 , wherein the consolidation includes performing a permutation of a plurality of different possible labels.

3 . The apparatus of claim 1 , wherein the consolidation includes comparing a distance.

4 . The apparatus of claim 3 , wherein the distance is determined based on a Hamming distance.

5 . The apparatus of claim 3 , wherein the distance is determined based on a measure of a minimum edit.

6 . The apparatus of claim 3 , wherein the processor is further configured to select a permutation based on the distance.

7 . The apparatus of claim 1 , wherein the data is clustered during a finite period of time.

8 . The apparatus of claim 1 , wherein the processor is further configured to receive an audio segment.

9 . The apparatus of claim 1 , wherein the processor is further configured to receive an audio segment.

10 . The apparatus of claim 1 , wherein the processor is further configured to store a plurality of audio segments.

11 . The apparatus of claim 1 , wherein clustering the data includes finding a cluster identifier (ID).

12 . The apparatus of claim 1 , wherein the processor is further configured to associate a cluster with an audio segment.

13 . The apparatus of claim 1 , wherein outputting the result includes outputting a label for a next occurring audio segment.

14 . A method of managing an audio stream, the method comprising:
clustering data to a point in time; consolidating the data with a past decision value; and outputting a result that minimizes a difference with the past decision value.

15 . The method of claim 14 , further comprising performing a permutation of a plurality of different possible labels.

16 . The method of claim 14 , further comprising comparing a distance to determine the consolidation.

17 . The method of claim 14 , further comprising determining the distance using a Hamming distance.

18 . The method of claim 17 , further comprising selecting a permutation based on the distance.

19 . The method of claim 14 , wherein the data is clustered during a finite period of time.

20 . A program product comprising
a computer readable storage medium having computer readable program code embodied therewith, the computer readable program code being executable by a processor to cluster the data to a point in time and consolidates the data with a past decision value, and to output a result that minimizes a difference with the past decision value.

Assignees

• IBM

Inventors

• Church Kenneth W
• Dimitriadis Dimitrios B
• Fousek Petr
• Pelecanos Jason W
• Zhu Weizhong

Classifications

• G10L15/063

  Training · CPC title

• G10L2015/0631

  Creating reference templates; Clustering · CPC title

• G10L15/04

  Segmentation; Word boundary detection · CPC title

• G10L15/083

  Recognition networks (G10L15/142, G10L15/16 take precedence) · CPC title

• G10L15/10 Primary

  using distance or distortion measures between unknown speech and reference templates · CPC title