Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
2002-01-30
2004-01-13
Knepper, David D. (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S243000
Reexamination Certificate
active
06678656
ABSTRACT:
TECHNICAL FIELD
This invention relates generally to speech recognition, and more particularly to distributed speech recognition.
BACKGROUND
Speech recognition techniques are known. Many speech recognition techniques provide for digitization of the speech information and subsequent processing to facilitate pattern matching that supports recognition of the speech information itself. Such processing often includes characterizing certain aspects of the speech information and representing those characterized aspects in some way, such as with cepstral coefficients. Generally speaking, the accuracy, speed, and reliability of a given speech recognition technique, using any given characterization approach, will improve as pattern matching resources increase. Unfortunately, for many applications, the results of the speech recognition activity are often required in small, portable user devices that have significantly limited resources. As a result, speech recognition for such devices often suffers for lack of such resources.
One proposed solution is to at least partially characterize the speech at the user device, and then provide that characterization information to a remote location (such as a speech recognition server) having significant resources. Those resources can then be used to complete the recognition process with presumed improved accuracy. One such distributed solution has been designed that will suitably process and characterize a voice signal within an 8 KHz frequency band, thereby providing a so-called telephone-band level of service. There are instances, however, when a wider bandwidth frequency band, such as a 16 KHz frequency band, would be desirable. Unfortunately, such an 8 KHz solution is not readily scalable to allow simple accommodation of an increased bandwidth signal. At the same time, however, many 8 KHz solutions are effective for their intended use and represent a desired embodiment for such applications.
One solution would be to simply provide a completely separate embodiment for dealing with larger bandwidth signals. This solution, however, requires completely parallel approaches that can necessitate a commensurate high level of resource dedication.
REFERENCES:
patent: 4630305 (1986-12-01), Borth et al.
patent: 5583961 (1996-12-01), Pawlewski et al.
patent: 5642383 (1997-06-01), Suzuki
patent: 6216103 (2001-04-01), Wu et al.
Lawrence Rabiner and Biing-Hwang Juang, Fundamentals of Speech Recognition, 1993 by AT&T, pp. 91-97.*
Sadaoki Furui and M. Mohan Sondhi, Advances in Speech Signal Processing, 1992 by Marcel Dekker, Inc., pp. 313-320.
Cheng Yan Ming
Macho Dusan
Fitch Even Tabin & Flannery
Knepper David D.
LandOfFree
Noise reduced speech recognition parameters does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Noise reduced speech recognition parameters, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Noise reduced speech recognition parameters will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3203862