Call classification by automatic recognition of speech

• US 20030088403A1
• Filed: 10/23/2001
• Published: 05/08/2003
• Est. Priority Date: 10/23/2001
• Status: Abandoned Application

• Current Assignee: Avaya Incorporated
• Original Assignee: Avaya Incorporated
• Inventors: Spencer, Douglas A., Das, Sharmistha Sarkar, Chan, Norman C., Wages, Danny M.
• Application Number:
  US10/037,590
• Publication Number:
  US 20030088403A1
• Time in Patent Office: Days
• US Class Current: 704/213
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

Classifying a call to a called destination endpoint by a call classifier. The call classifier is responsive to information received from the called destination endpoint to perform the call classification.

Claims

1. A method for doing call classification on a call to a destination endpoint, comprising the steps of:

receiving audio information from the destination endpoint;

analyzing received audio information for words using automatic speech recognition; and

determining the call classification from the analyzed words.

2. The method of claim 1 wherein the analyzed words are formed as phrases.

3. The method of claim 1 wherein the step of analyzing comprises performing front-end feature extraction on the received audio information to produce a full feature vector.

4. The method of claim 3 wherein the step of analyzing further comprises computing log likelihood probability from the full feature vector.

5. The method of claim 4 wherein the step of analyzing further comprises updating a dynamic programming network used in the step of analyzing in response to the computed log likelihood probability.

6. The method of claim 5 wherein the step of updating comprises the step of executing an Viterbi process.

7. The method of claim 5 further comprises the step of pruning the nodes in the dynamic programming network used in the step of analyzing.

8. The method of claim 7 further comprises the step of expanding a grammar network used in the step of analyzing.

9. The method of claim 8 further comprises the step of performing grammar backtracking in response to the expanded grammar network.

10. The method of claim 9 wherein the step of backtracking comprises the step of executing another Viterbi process.

11. The method of claim 1 wherein the step of determining comprises executing an inference engine in response to analyzed words.

12. The method of claim 11 further comprises the step of analyzing the audio information to detect tones;

and

the step of determining further responsive to the detection of tones for determining the call classification.

13. The method of claim 12 further comprises the step of analyzing the audio information to identify energy in the audio information;

and

the step of determining further responsive to the identification of energy for determining the call classification.

14. The method of claim 13 further comprises the step of analyzing the audio information to identify zero crossings in the audio information;

and

the step of determining further responsive to the identification of zero crossings for determining the call classification.

15. A method for doing call classification on a call to a destination endpoint, comprising the steps of:

receiving audio information from the destination endpoint;

analyzing received audio information for a first classification;

analyzing received audio information using automatic speech recognition for a second classification; and

determining the call classification from the first classification and the second classification.

16. The method of claim 15 wherein the first classification is one of tone detection, energy analysis, or zero crossing analysis.

17. The method of claim 16 further comprises the step of analyzing for a third classification;

and

the step of determining further responsive to the third classification.

18. The method of claim 17 wherein the third classification is one of tone detection, energy analysis, or zero crossing analysis.

19. The method of claim 18 wherein the step of determining comprises executing an inference engine.

20. The method of claim 19 wherein the step of analyzing received audio information using automatic speech recognition comprises the step of executing a Hidden Markov Model.

21. An apparatus for classifying a call to a called destination endpoint, comprising:

a receiver for receiving audio information from the called destination endpoint;

automatic speech recognizer for determining words in the received audio information; and

an inference engine for classifying the call destination endpoint in response to the determined words.

22. The apparatus of claim 21 wherein the determined words are formed as phrases.

23. The apparatus of claim 21 further comprises an analyzer for determining another classification from the received audio information.

24. The apparatus of claim 23 wherein the analyzer is one of a tone detector, energy detector, or a zero crossings detector.

25. The apparatus of claim 24 wherein the automatic speech recognizer is executing a Hidden Markov Model.