Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Patent
1995-12-12
1999-03-16
Hudspeth, David R.
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
704211, 704243, 704276, G10L 506
Patent
active
058842608
DESCRIPTION:
BRIEF SUMMARY
The present invention relates to a method and system for signal processing, by which method and system features representing distinct sound pictures in auditory signals are extracted from transients in the auditory signals. The result of the processing may be used for identification of sound or speech signals or for quality measurement of audio products or systems, such as loudspeakers, hearing aids, telecommunication systems, or for quality measurement of acoustic conditions. The method of the present invention may also be used in connection with speech compression and decompression in narrow band telecommunication.
In the prior art methods of signal analysing of auditory signals, the signals are considered to be steady state over a short time of period, and a form of short time spectral analysis is used under this assumption.
The human ear has the ability to simultaneously catch fast sound signals, detect sound frequencies with great accuracy and differentiate between sound signals in complicated sound environments. For instance it is possible to understand what a singer is singing in an accompaniment of musical instruments.
In prior art methods of signal analysis and in the method of the present invention it is assumed that the cochlea in the human ear can be regarded as an infinite number of bandpass filters, IBP, within the frequency range of the human ear.
The time response f(t) for one bandpass filter due to an excitation can be separated into two components, the transient response, ft(t), and the steady state response, fs(t),
Traditional signal processing is based on the steady state response fs(t), and the transient response ft(t) is assumed to vanish very fast and to be without importance for the perception, see for example "Principles of Circuit Synthesis", McGraw-Hill 1959, Ernest 5. Kuh and Donald O. Pederson, page 12, lines 9-15, where it is stated that: initial state of the network is ignored".
Thus, when students are introduced to the world of signal analysis, they learn at a very early stage that the transient response, i.e. the response due to the initial state of the network should be ignored because it vanishes within a very short period of time. Furthermore, it is rather difficult to analyse these transient signals by use of traditional linear methods of analysis.
The ability of the human ear to hear very short sounds and at the same time detect frequencies with great accuracy is in conflict with the traditional filterbased spectrum analysis. The time window (twice the rise time) of a bandpass filter is inversely proportional to the bandwidth, frequency.
Thus, if a rise time of 5 ms is required the consequence is that the frequency resolution is no better than 400 Hz.
As the detection of these transients is in conflict with a high frequency resolution, the detecting by the human ear of these transients must take place in an alternative manner. It has not been examined how the human ear is able to detect these signals, but it might be possible that the cochlea, when no sounds are received, is in a position of rest, where the cochlea will be very broad-banded. When a sound signal is received, the cochlea may start to lock itself to the frequency component or components within the signal. Thus, the cochlea may be broad-banded in its starting position, but if one or more stable frequencies are received the cochlea may lock itself to this frequency or these frequencies with a high accuracy.
Today it is known that the nerve pulses launched from the cochlea are synchronized to the frequency of a tone if the frequency is less than about 1.4 kHz. If the frequency is higher than 1.4 kHz the pulses are launched randomly and less than once per cycle of the frequency.
Signal analysis based on filter bank spectrum analysis is disclosed in GB 2213623 which describes a system for phoneme recognition. This system comprises detecting means for detecting transient parts of a voice signal, where the principal object of the transient detection is the detection of a point where the speech spectrum varies mo
REFERENCES:
patent: 3838217 (1974-09-01), Dreyfus
patent: 3855418 (1974-12-01), Fuller
patent: 3919479 (1975-11-01), Moon
patent: 3940565 (1976-02-01), Lindenberg
patent: 3946157 (1976-03-01), Dreyfus
patent: 4701953 (1987-10-01), White
patent: 4817155 (1989-03-01), Briar et al.
patent: 4817159 (1989-03-01), Hoshimi et al.
patent: 4821325 (1989-04-01), Martin et al.
patent: 5035242 (1991-07-01), Franklin et al.
patent: 5091948 (1992-02-01), Kametani
patent: 5220639 (1993-06-01), Lee
patent: 5327518 (1994-07-01), George et al.
J. N. Holmes, "Speech Synthesis and Recognition," Chapman and Hall, London, 1988, pp. 54-60 and 71.
"Microsoft Press Computer Dictionary," second ed., 1994, p.345.
"Principles of circuit Synthesis", McGraw-Hill 1959, Ernest S. Kuh and Donald O. Pederson. pp. 6-21.
Hudspeth David R.
Storm Donald L.
LandOfFree
Method and system for detecting and generating transient conditi does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for detecting and generating transient conditi, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for detecting and generating transient conditi will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-827302