Method and system for estimation of a source of a voice signal

• US 6,208,970 B1
• Filed: 12/21/1998
• Issued: 03/27/2001
• Est. Priority Date: 12/21/1998
• Status: Expired due to Term

• Current Assignee: Avaya Incorporated
• Original Assignee: Nortel Networks Limited (Nortel Networks Corporation)
• Inventors: Ramanan, Kandiah
• Primary Examiner(s): Hudspeth, David
• Assistant Examiner(s): Azad, Abul K.
• Application Number:
  US09/216,971
• Time in Patent Office: 827 Days
• Field of Search: 704/270, 704/231, 704/251, 704/253, 379/80, 379/88.01, 379/88.12, 379/88.19, 379/88.22
• US Class Current: 704/270
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: H04M 2203/2027 Live party detection
• CPC Class Codes: H04M 3/5158 in combination with automated outdialling systems devices for signalling identity of wanted subscriber H04M1/26

Abstract

Methods of estimating if a voice signal is more likely formed by a machine or human, are disclosed. The methods are particularly well suited for use in an automated call establishment system. A voice signal is received at a computing device. Speech recognition steps are applied to the voice signal to recognize spoken words in the signal. Recognized words are compared to words likely generated by a machine or a live human. Based on the comparison, an estimate is made that the voice signal more likely originates with a machine or a human. Conventional call progress tone recognition may be combined with speech recognition to obtain an improved estimate of call progress. As a result of the methods, calls to machines may be disconnected or connected to a pre-recorded message, while calls to humans may be connected to call handling agents or a different pre-recorded messages. An automated call establishment system using the methods is also disclosed.

Claims

1. A method of operating an automated call establishment system including a computing device, said method comprising:

a. establishing a call connection with a far-end;

b. receiving at said computing device a voice signal originating with a voice source at said far-end;

c. applying speech recognition to said voice signal at said computing device to form indicators of recognized spoken words in said voice signal;

d. comparing said formed indicators to one or more stored indicators stored at said computing device identifying words likely generated by one of a machine and a human voice source;

e. classifying said voice signal as likely originating with one of a machine voice source and a human voice source, based on said comparing.

2. The method of claim 1, further comprising:

receiving at least one tone signal, indicative of call progress to determine if said call connection has been established.

3. The method of claim 1, further comprising:

receiving at least one out-of band telephony signal, indicative of call progress to determine if said call connection has been established.

4. The method of claim 3, wherein said out-of band telephony signal is one of an SS7 and ISDN signal.

5. The method of claim 1, further comprising:

counting words in said voice signal prior to a significant pause, and classifying said voice signal as originating with one of a machine voice source and a human voice source, based on said number of counted words prior to said significant pause.

6. The method of claim 5, wherein said voice source is classified as a machine voice source if said number of counted words prior to said pause exceeds a defined threshold number of words.

7. The method of claim 1, further comprising detecting noise characteristic of a magnetic audio tape in said voice signal, and classifying said voice signal as originating with a machine, in response to detecting said noise characteristic of a magnetic audio tape.

8. The method of claim 1, further comprising:

disconnecting said call connection if said voice source is classified as a machine voice source.

9. The method of claim 1, further comprising:

transferring said call connection to a call handling agent if said voice source is classified as a human voice source.

10. The method of claim 1, further comprising:

playing a first recorded message over said call connection, if said voice source is classified as a machine voice source.

11. The method of claim 1, further comprising:

playing a second recorded message over said call connection, if said voice source is classified as a human voice source.

12. A method of estimating if a voice signal is more likely formed by a machine or human, comprising:

a. receiving said voice signal at a computing device;

b. applying speech recognition to said voice signal at said computing device to form indicators of recognized spoken words in said voice signal;

c. comparing said formed indicators to one or more indicators stored at said computing device identifying words likely generated by one of a machine and a human;

d. estimating if said voice signal is more likely formed by a machine or a human based on said comparing.

13. The method of claim 12, wherein said voice signal is received as a result of establishing a call connection with a far end voice source.

14. The method of claim 13, further comprising:

counting words in said voice signal prior to a significant pause, and wherein said estimating is based on said number of counted words prior to said significant pause.

15. A call establishment system comprising:

a network interface for connection with a communications network;

a processor in communication with said interface;

processor readable memory in communication with said processor, said memory storing instructions adapting said system to;

a. establish an outgoing call using said network interface;

b. receive a voice signal from a far-end of said outgoing call;

c. apply speech recognition to said voice signal to form indicators of recognized spoken words in said voice signal;

d. compare said formed indicators to one or more stored indicators identifying words likely generated by one of a machine and a human voice source;

e. classify said voice source as a machine or human voice source.

16. The call establishment system of claim 15, further comprising an input device for modifying said stored indicators.

17. The call establishment system of claim 15, wherein said memory further stores processor readable instructions to while establishing a call, detect at least one tone signal indicative of call progress to determine if a call connection has been established.

18. A call establishment system comprising:

means for establishing an outgoing call;

means for recognizing speech in a voice signal at a far end of said outgoing call;

means for comparing said recognized speech to words likely generated by one of a machine and a human voice source;

means for classifying said voice source as a machine or human voice source in communication with said means for comparing.

19. The system of claim 18, further comprising:

means for detecting noise characteristic of a magnetic audio tape in said voice signal, and

means for classifying said voice signal as originating with a machine, in response to said means for detecting, detecting noise characteristic of a magnetic audio tape.

20. A method of operating an automated call establishment system, including a computing device, said method comprising:

a. establishing a call connection with a far-end;

b. receiving a voice signal from said far-end at a computing device;

c. recognizing words within said voice signal using speech recognition;

d. counting a number of recognized words in said voice signal prior to a significant pause in said voice signal;

e. classifying said voice signal as likely originating with one of a machine voice source or a human voice source, based on said number of recognized words.