Telephonic communications – Audio message storage – retrieval – or synthesis – Voice activation or recognition
Reexamination Certificate
1998-01-16
2001-04-17
Tsang, Fan (Department: 2645)
Telephonic communications
Audio message storage, retrieval, or synthesis
Voice activation or recognition
C379S067100, C379S080000, C379S088220, C379S093030
Reexamination Certificate
active
06219407
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to speech/speaker recognition and telephone mail messaging and, more particularly, to apparatuses and methods for improved digit recognition and/or caller identification utilizing speech/speaker recognition in telephone mail messaging.
Typically, in a telephone system having a voice mail feature, a caller leaves a telephone message which may include his name, telephone number and/or a brief request or message on a receiving party's voice mail equipment. As is known, the telephone number that is left usually informs the voice mail user where the caller may be reached over the telephone. Conventional automatic speech recognition (ASR) decoding may provide the user with a decoded text representation of the phone message. However, an error in decoding even one digit of the telephone number of the caller can make an entire telephone message useless since a user may not be able to return a call (unless, of course, the user plays back a recorded representation of the phone message).
There exist telephone devices (and services) that allow a receiving party to trace back or record a telephone number of a telephone set from which a caller placed a call. However, this is not always useful since a caller may have called from some temporal location (e.g., a street phone) or may have left a telephone number to call back that is different from a telephone number at his current location. Furthermore, user identification alone does not help to identify the phone number to call back since the caller may have many phone numbers where he can be reached, e.g., home, office, hotels during his travels, etc.
In addition to voice mail messaging systems, a fast growing area in the consumer communications market is text-independent speaker recognition as disclosed in U.S Ser. No. 08/788,471 filed on Jan. 28, 1997, entitled: “Text-independent Speaker Recognition for Command Disambiguity and Continuous Access Control”. It is known that a problem with text-independent speaker recognition is that a textual context, in general, is difficult to use to improve the accuracy of speaker recognition. Also, with regard to telephone applications, since the bandwidth associated with a typical telephone line may tend to reduce the accuracy associated with ASR, telephone continuous speaker-independent recognition decoding has been considered to be a challenging task. Especially with the additional difficulties of microphone mismatch (e.g., speaker phone, cellular phones, carbon and/or electric microphones) and channel variability (e.g., from one phone call to another, the path through the telephone network can vary dramatically, which in turn has a severe effect on the distortions and signature introduced by the channel).
It would be highly desirable and advantageous to provide apparatuses and methods which overcome the drawbacks and limitations described above with respect to ASR decoding of telephone voice mail messaging as well as telephone continuous speaker-independent recognition decoding.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide apparatuses and methods for performing speaker recognition in order to reduce a level of confusion and/or error in automatic speech, recognition (ASR) of digits of a telephone number left by a caller in a voice mail phone at message.
It is another object of the present invention to provide apparatuses and methods for utilizing automatic speech recognition of digits of a telephone number left by a caller in a voice message to improve automatic speaker recognition of the caller's identity.
In one aspect of the present invention, a telephone voice mail messaging system for performing recognition of characters included in a telephone message left by a caller comprises: speaker recognition means, responsive to the telephone message, for extracting a voice model of the caller from the telephone message and comparing the voice model of the caller against a plurality of pre-obtained voice models respectively associated with a plurality of potential callers to find a substantial match between the voice model of the caller and one of the voice models associated with one of the potential callers and, if a substantial match is found, then the speaker recognition means identifying at least one pre-obtained set of characters attributed to the potential caller associated with the substantially matching voice model; character spotting means, responsive to the telephone message, for identifying sets of characters in the telephone message; and comparator means for comparing the characters of the at least one pre-obtained set of characters identified by the speaker recognition means against the characters of a set of characters identified by the character spotting means and, if the characters of the sets substantially match, then outputting the characters of the at least one pre-obtained set of characters as representing the set of characters left by the caller as part of the telephone message. Such character set may preferably be combined with a decoded representation of the telephone message to form a corrected decoded message, which may be provided to a user interface unit (preferably, a CRT display terminal) so that a voice mail user may learn who called him, what the return telephone number is, and any request or message left. The user interface unit may also preferably provide the user with the ability to selectively place a return call, read the message or play back an audio representation of the message.
It is to be appreciated that sets of characters may include: both numbers and letters, as in addresses; numbers only, as in telephone numbers and social security numbers; or letters only, such as in names. Such examples are only exemplary and thus other character sets may be employed.
In another aspect of the present invention, a telephone voice mail messaging system for performing caller identification with respect to a telephone message left by a caller comprises: character spotting means, responsive to the telephone message, for identifying sets of characters in the telephone message and comparing an identified set of characters against a plurality of pre-obtained sets of characters associated with a plurality of potential callers to find a substantial match between the identified set of characters and at least one of the pre-obtained sets of characters and, if a substantial match is found, then the character spotting means further identifying a list of potential callers attributed to the at least one substantially matching pre-obtained set of characters; and speaker identification means, responsive to the telephone message, for extracting a voice model of the caller from the telephone message and comparing the voice model of the caller against respective voice models associated with the potential callers on the list to find a substantial match between the voice model of the caller and the voice model of one of the potential callers and, if a substantial match is found, then the speaker identification means identifying the potential caller associated with the substantially matching voice model. The identity of the caller may preferably be provided to a service access server which may selectively permit the user access to a service. Still further, the identity of the caller may preferably be provided to a user interface unit with similar capabilities as mentioned above and described herein.
These and other objects, features and advantages of the present invention will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in connection with the accompanying drawings.
REFERENCES:
patent: 4935954 (1990-06-01), Thompson et al.
patent: 4942598 (1990-07-01), Davis
patent: 5199080 (1993-03-01), Kimura et al.
patent: 5465401 (1995-11-01), Thompson
patent: 5479490 (1995-12-01), Nakashima
patent: 5568540 (1996-10-01), Greco et al.
patent: 5572578 (1996-11-01), Lin et al.
patent: 5594789 (1997-01-01), Seazholtz et al.
patent: 5604790 (1997-02-0
Kanevsky Dimitri
Maes Stephane Herman
F. Chau & Associates LLP
Hoosain Allan
International Business Machines - Corporation
Tsang Fan
LandOfFree
Apparatus and method for improved digit recognition and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus and method for improved digit recognition and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus and method for improved digit recognition and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2484553