Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1999-09-30
2003-05-06
Chawan, Vijay (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S239000, C704S221000, C704S237000, C704S243000
Reexamination Certificate
active
06560575
ABSTRACT:
BACKGROUND OF THE INVENTION
Field of the Invention
The present invention relates to a speech processing apparatus and method. The invention has particular, although not exclusive relevance to the masking of noise in an input signal, such as an input speech signal.
In some speech recognition and speech verification systems where there can be high levels of noise or where the noise level can change considerably, mis-recognition and mis-verification can result due to the energy in the noise signal at some frequencies being greater than the energy of the input speech at those frequencies. U.S. Pat. No. 4,918,732 addresses this problem and alleviates it by masking out the frequencies in the speech signal which may have an energy below the energy of the background noise, both during training and during subsequent recognition or verification, so that these portions are not taken into consideration during the matching process. The system described in U.S. Pat. No. 4,918,732 assumes a constant noise level in each frame of the input speech signal and can not be used, therefore, if an automatic gain controller is used, since the gain applied to each frame of the input speech signal will be different.
SUMMARY OF THE INVENTION
The present invention provides a consistency checking apparatus for checking the consistency between a first sequence of frames representative of a first signal and a second sequence of frames representative of a second signal using a matching score and the results of a matching process performed on the first and second sequences of frames, the apparatus comprising: means for determining an average frame score by dividing the matching score by the number of frames in the first signal which are matched with the frames in the second signal; means for determining the score of a worst matching portion between the first and second signals; memory means for storing data defining a model of consistent training examples; means for comparing the average frame score and the score of the worst matching portion with said stored model; and means for determining whether or not the first and second input signals are consistent from the output of said comparing means.
REFERENCES:
patent: 5339385 (1994-08-01), Higgins
patent: 5625749 (1997-04-01), Goldenthal et al.
patent: 5684925 (1997-11-01), Morin et al.
patent: 5710866 (1998-01-01), Alleva et al.
patent: 5729656 (1998-03-01), Nahamoo et al.
patent: 5737722 (1998-04-01), Kopp et al.
patent: 5907824 (1999-05-01), Tzirkel-Hancock
patent: 5956678 (1999-09-01), Hab-Umbach et al.
patent: 5960393 (1999-09-01), Cohrs et al.
patent: 5960395 (1999-09-01), Tzirkel-Hancock
patent: 6012027 (2000-01-01), Bossemeyer, Jr.
patent: 6223155 (2001-04-01), Bayya
patent: 6226610 (2001-05-01), Keiller et al.
patent: 6240389 (2001-05-01), Keiller et al.
patent: 19804047 (1999-08-01), None
patent: 0 789 349 (1997-08-01), None
Pal, et al., Effect of Wrong Samples On The Convergence of Learning Process, Information Science, Mar. 1992, vol. 60, No. 1-2, pp. 77-105.
Bradshaw, et al., “A Comparison of Learning Techniques In Speech Recognition,” International Conference on Acoustics, IEEE, vol. CONF. 7, May 3, 1982, pp. 554-557.
LandOfFree
Speech processing apparatus and method 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 processing apparatus and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Speech processing apparatus and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3020759