Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1998-11-11
2004-04-06
McFadden, Susan (Department: 2655)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S226000
Reexamination Certificate
active
06718301
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to signal processing, and more particularly to a system for measuring speech content in sound.
BACKGROUND
A difficult problem for scientists in voice recognition is to electronically differentiate speech-like sound from other sounds. Speech-like sound is sound with a time-frequency description that changes like that of speech sounds. The human brain is very capable of recognizing the difference between speech-like sounds and other sounds. For example, humans can easily differentiate between the sibilant whirring of computer fan noise and a person talking. However, it is extremely complicated to produce electronics which can tell the difference between noises and speech-like sound. One of the complications is that the noises we hear have energy in much the same frequencies as speech-like sound. Another complication is that some vocalizations are not simple talking and therefore may have sound characteristics which are closer to non-speech sound than speech-like sound. Yet another complication is that there are a variety of different speech-like sounds which demonstrate substantially different characteristics. For example, the characteristics of simple talking compared to singing are readily distinguishable to the human ear, yet such differences may confuse systems attempting to electronically differentiate speech-like sound from other sounds.
Thus, there is a need in the art for a system which differentiates speech-like sound from other sounds. The system should be able to characterize vocalizations which are not simple talking. Furthermore, the system should be useful for differentiating a variety of different kinds of speech-like sounds from other sounds.
SUMMARY OF THE INVENTION
Upon reading and understanding the present disclosure it is recognized that the inventive subject matter described herein satisfies the foregoing needs in the art and several other needs in the art not expressly noted herein. The following summary is provided to give the reader a summary which is not intended to be exhaustive or limiting and the scope of the invention is provided by the attached claims and the equivalents thereof.
One embodiment of the present invention provides a method and apparatus for a system for measuring speech content in sound. In one embodiment a method is provided for receiving an input signal; extracting a signal related to a time-dependent power of the input signal; determining a time-dependent mean of the signal, M; determining a time-dependent deviation of the signal from the mean, D; and estimating a time-dependent speech-to-noise ratio from M and D. In one embodiment, the extracted signal is an envelope produced using a non-negative function of the input signal. In one embodiment, the estimating includes comparing M and D to a predetermined mapping of a relationship between M and D to speech-to-noise ratio to obtain an estimated speech-to-noise ratio. Embodiments in which the deviation D is the standard deviation are demonstrated. Other deviations are demonstrated. Various analog and digital embodiments are demonstrated herein. Single band and multiple band embodiments are provided. Various filtering systems are demonstrated, including recursive and nonrecursive. Multiple signal extraction methods are demonstrated. Uses of the estimated time-dependent speech-to-noise ratios are demonstrated.
In one embodiment apparatus and process are provided relating to a system receiving an audio signal; converting the audio signal to an electrical signal; converting the electrical signal to a digital representation; bandpass filtering the signal using one or more digital filters to produce a plurality of filtered digital signals; for each digital signal of the plurality of filtered digital signals: extracting an envelope related to a time-dependent power of the signal; determining a time-dependent mean of the envelope, M; determining a time-dependent deviation of the envelope from the mean, D; and estimating a time-dependent speech-to-noise ratio from M and D using a predetermined mapping or relationship between M and D to speech-to-noise ratio; producing a processed digital signal using the plurality of filtered digital signals and their respective estimated time-dependent speech-to-noise ratios; and converting the processed digital signal to a processed analog signal. Alternate embodiments are provided to demonstrate the subject matter of the present patent application. Several applications of the present subject matter are discussed.
REFERENCES:
patent: 3527901 (1970-09-01), Geib
patent: 4074264 (1978-02-01), Wilmot
patent: 4366349 (1982-12-01), Adelman
patent: 4396806 (1983-08-01), Anderson
patent: 4419544 (1983-12-01), Adelman
patent: 4471490 (1984-09-01), Bellafiore
patent: 4627103 (1986-12-01), Fukuhara
patent: 4628529 (1986-12-01), Borth et al.
patent: 4630304 (1986-12-01), Borth et al.
patent: 4637402 (1987-01-01), Adelman
patent: 4811404 (1989-03-01), Vilmur et al.
patent: 4852175 (1989-07-01), Kates
patent: 4882762 (1989-11-01), Waldhauer
patent: 4887299 (1989-12-01), Cummins et al.
patent: 5012519 (1991-04-01), Adlersberg et al.
patent: 5185168 (1993-02-01), Takahashi
patent: 5267322 (1993-11-01), Smith et al.
patent: 5323337 (1994-06-01), Wilson et al.
patent: 5390254 (1995-02-01), Adelman
patent: 5434924 (1995-07-01), Jampolsky
patent: 5502769 (1996-03-01), Gilbertson
patent: 5544250 (1996-08-01), Urbanski
patent: 5550924 (1996-08-01), Helf et al.
patent: 5553152 (1996-09-01), Newton
patent: 5581747 (1996-12-01), Anderson
patent: 5659621 (1997-08-01), Newton
patent: 5687241 (1997-11-01), Ludvigsen
patent: 5757933 (1998-05-01), Preves et al.
patent: 5822442 (1998-10-01), Agnew et al.
patent: 5825631 (1998-10-01), Prchal
patent: 5835611 (1998-11-01), Kaiser et al.
patent: 5852668 (1998-12-01), Ishige et al.
patent: 5862238 (1999-01-01), Agnew et al.
patent: 5884255 (1999-03-01), Cox
patent: 6041129 (2000-03-01), Adelman
patent: 6175634 (2001-01-01), Graumann
patent: 6236731 (2001-05-01), Brennan et al.
patent: 6240192 (2001-05-01), Brennan et al.
patent: 6347148 (2002-02-01), Brennan et al.
patent: 6366863 (2002-04-01), Bye et al.
patent: 6389142 (2002-05-01), Hagen et al.
patent: 6449662 (2002-09-01), Armitage
patent: 2001/0007050 (2001-07-01), Adelman
patent: 2002/0076073 (2002-06-01), Taenzer
patent: WO-00/21332 (2000-04-01), None
Mischa Schwartz, William R. Bennett, and Seymour Stein, Communication Systems and Techniques, McGraw-Hill, 1966, p. 30, 36, and 103.*
Griffing, Terry S., et al., “Acoustical Efficiency of Canal ITE Aids”,Audecibel, (Spring 1983), 30-31.
Griffing, Terry S., et al., “Custom canal and mini in-the-ear hearing aids”,Hearing Instruments, vol. 34, No. 2, (Feb. 1983), 31-32.
Griffing, Terry S., et al., “How to evaluate, sell, fit and modify canal aids”,Hearing Instruments, vol. 35, No. 2, (Feb. 1984), 3.
Mahon, William J., “Hearing Aids Get a Presidential Endorsement”,The Hearing Journal,,(Oct. 1983), 7-8.
Sullivan, Roy F., “Custom canal and concha hearing instruments: A real ear comparison”,Hearing Instruments, vol. 40, No. 4, (Jul. 1989), 5.
Sullivan, Roy F., “Custom canal and concha hearing instruments: A real ear comparison Part II”,Hearing Instruments, vol. 40, No. 7, (Jul. 1989),6.
Elberling, C., et al., “The Design and Testing of a Noise Reduction Algorithm Based on Spectral Subtraction”,Scand Audiol, Suppl. 38, 39-49, (1993).
Ludvigsen, C., “Comparison of Certain Measures of Speech and Noise Level”,Scand Audiol, 21, 23-29, (1992).
McFadden Susan
Schwegman Lundberg Woessner & Kluth P.A.
Starkey Laboratories, Inc.
LandOfFree
System for measuring speech content in sound does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for measuring speech content in sound, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for measuring speech content in sound will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3240852