Filter bank determination of discrete tone frequencies

Details Filter bank determination of discrete tone frequencies Filter bank determination of discrete tone frequencies

1994-06-14

1997-03-25

MacDonald, Allen R.

395 216, 395 214, G10L 302, G10L 900

Patent

active

056153023

DESCRIPTION:

BRIEF SUMMARY
BACKGROUND OF THE INVENTION

1. Field of the Invention
The invention relates to methods and apparatus for extracting the information content of audio signals, in particular audio signals associated with human speech.
2. Related Art
Conventional devices for extracting the information content from human speech are plagued with difficulties. Such devices, which include voice activated machines, computers and typewriters, typically seek to recognize, understand and/or respond to spoken language. Speech compressors seek to minimize the number of data bits required to encode digitized speech in order to minimize the cost of transmitting such speech over digital communication links. Hearing-aids seek to augment the hearing impaired's ability to extract information from speech and thus better understand conversations. Numerous other speech interpreting or responsive devices also exist.
As disclosed herein, the difficulties encountered by these devices and their resulting poor performance stem from the fact that they incorporate principles of operation that are wholly unlike the operating principles of the human ear. Since such devices fail to incorporate an information extraction principle similar to that found in the ear, they are incapable of extracting and representing speech information in an efficient manner.
Chappell in "Filter Technique Offers Advantages for Instantaneous Frequency Measurement" published in Microwave System News and Communications Technology, June, 1986, discloses the basic concept of channelized filter discriminators or ratio detectors. Chappell applies the technique to measuring the frequency of individual radar pulses rather than speech and does not address measurement of combination of harmonics for frequency diversity processing. In addition, Chappell uses butterworth filters with a non-linear frequency discriminator curve rather than Gaussian filters, as is disclosed herein, with a perfectly linear discriminator curve or Gaussian/exact log discriminator curve.
Morlet, et al., in "Wavelet Propagation and Sampling Theory" published in Geophysics in 1982, discloses a filter bank with Gaussian filters equally spaced along a logarithmic frequency axis. The system is applied to seismic waves, rather than speech and does not address the measurement and combination of harmonics for frequency diversity processing.
Hartman in "Hearing a Mistuned Harmonic in an Otherwise Periodic Complex Tone", published in 1990 in the Journal of the Acoustical Society of America, and in Chapter 21 of Auditory Function "Pitch Perception and the Segregation and Integration of Auditory Entities" describes the abilities of the auditory system to recognize and distinguish different sounds, but not how this is accomplished. The use a frequency discrimination process to measure harmonic frequencies and "pitch meter" that fits harmonic templates to resolve frequency components using conventional spectral analysis, is also disclosed. However, none of these references can account for observed functional behavior of the human ear. In addition, none of the references discloses that the ear is primarily a modulation detector rather than a general purpose sound detector, speech modulation uses a hybrid AM/FM signaling scheme with frequency diversity via harmonically related carriers. The reasons why ones perception of pitch is logarithmic is that proper FM demodulation of harmonics requires band pass filters with band widths proportional to their center frequencies in a logarithmic relationship. Finally, there is no disclosure of a ratio detector.
Information encoded in signals can be extracted in numerous ways. Usually, the optimal way to extract information from signals is to employ the same approach used for encoding the information. The human ear does not appear to employ conventional data processing methods of extracting information from sound signals, such as methods using Fourier coefficients, Wavelet transform coefficients, linear prediction coefficients or other common techniques dependent on measurements of the sou

REFERENCES:
patent: 3755627 (1973-08-01), Berkowitz et al.
patent: 3801983 (1974-04-01), Woolley
patent: 3833767 (1974-09-01), Wolf
patent: 3989896 (1976-11-01), Reitboeck
patent: 4001702 (1977-01-01), Kaufman
patent: 4665390 (1987-05-01), Kern et al.
patent: 5214708 (1993-05-01), McEachern
Geckinli, et al., "Speech Synthesis Using AM/FM Sinusoids and Band-Pass Noise", Signal Processing 8 (1985), pp. 339-361.

Affiliated with

Also associated with

Rating

Filter bank determination of discrete tone frequencies does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Filter bank determination of discrete tone frequencies, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Filter bank determination of discrete tone frequencies will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2211095