Patent
1994-06-13
1996-01-23
MacDonald, Allen R.
395 24, 395 231, G10L 900
Patent
active
054871298
DESCRIPTION:
BRIEF SUMMARY
FIELD OF THE INVENTION
The present invention relates generally to pattern matching in noisy environments and to speech pattern matching in noisy environments in particular.
DESCRIPTION OF THE PRIOR ART
Speech pattern matching is a known process in which an incoming test speech segment, such as a speech utterance, is compared to a collection of reference speech segments in order to find the reference speech segment in the collection that is most similar to the test speech segment. Similarity is defined by a score given to each reference segment with respect to the input test speech segment. The reference and test speech segments can each be represented by a set of features or by a model.
In the speech recognition task, the reference and test speech segments are uttered words and the collection of reference segments, known as templates, constitutes a pre-defined dictionary. In the speaker identification task, the reference segments are representative of voices of different people. In speech coding, such as through Vector Quantization (VQ), the test and reference segments are usually arbitrarily short segments and the VQ method represents each test segment by the index of the reference segment which is closest to it.
The matching capability of conventional algorithms deteriorates greatly in the presence of noise in the input speech. One approach to solving this problem is by speech enhancement preprocessing, a process which is reviewed in the book Speech Enhancement, edited by J. S. Lim, and published by Prentice-Hall, New-York, 1983. Application of such methods to speech recognition in a noisy environment in a car is described in the article by N. Dal Degan and C. Prati, "Acoustic Noise Analysis and Speech Enhancement Techniques for Mobile Radio Applications", Signal Processing, Vol. 15, pp. 43-56, 1988.
Some methods have been described which perform speech matching in a white noise environment for the purpose of speech recognition. Among them are the Short-time Modified Coherence (SMC) representation of speech, as described by D. Mansour and B. H. Juang, in their article, "The Short-Time Modified Coherence Representation and Noisy Speech Recognition", IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. ASSP-37, pp. 795-804, June 1989.
Other methods use noise robust distortion measures, such as a projection distortion measure or a Weighted Likelihood Ratio (WLR). Methods using the projection distortion method are discussed in the article by D. Mansour and B. H. Juang, "A Family of Distortion Measures Based Upon Projection Operation for Robust Speech Recognition", IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. ASSP-37, pp. 1659-1671, November 1989.
Unfortunately, the abovementioned methods fail when the noise is colored, as in the environment of a driving car.
Y. Ephraim, J. G. Wilpon and L. R. Rabiner, in the article, "A Linear Predictive Front-end Processor for Speech Recognition in Noisy Environments", International Conference on Acoustics, Speech and Signal Processing, ICASSP-87, pp. 1324--1327, Dallas Tex., 1987, present a method for speech recognition suitable for colored noise. In this method, the power spectrum of the noise is used, in an iterative algorithm, to estimate the Linear Prediction Coefficients (LPC) of clean speech from its noisy version. This algorithm requires extensive computations.
This last method and the SMC method were applied to speech recognition in car noise by I. Lecomte, M. Lever, J. Boudy and A. Tassy. Their results are discussed in the article, "Car Noise Processing for Speech Input", International Conference on Acoustics, Speech and Signal Processing, ICASSP-89, pp. 512-515. Glasgow UK, 1989.
SUMMARY OF THE INVENTION
It is an object of the present invention to overcome the problems of the prior art and to provide a pattern matching system which is operative in the presence of colored and quasi-stationary noise and which is computationally efficient. The present invention may be used, for example, for speech recognition, speaker ident
REFERENCES:
patent: 5046099 (1991-09-01), Nishimura
patent: 5319736 (1994-06-01), Hunt
Flomen Felix
Haimi-Cohen Raziel
Paiss Omri
Shallom Ilan D.
Doerrler Michelle
MacDonald Allen R.
The DSP Group
LandOfFree
Speech pattern matching in non-white noise does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Speech pattern matching in non-white noise, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Speech pattern matching in non-white noise will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1510526