Data processing: structural design – modeling – simulation – and em – Modeling by mathematical expression
Reexamination Certificate
2000-05-05
2004-06-15
Jones, Hugh (Department: 2123)
Data processing: structural design, modeling, simulation, and em
Modeling by mathematical expression
C340S601000
Reexamination Certificate
active
06751580
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to weather tracking systems and, more particularly, to such systems employing an acoustic signature matching device for tracking rapidly moving extreme weather systems such as tornadoes.
2. Description of Related Art
The recognition of acoustic signatures is a subset of the general problem of signal processing. Concomitant with developments in CPU power and memory size, software-based signal processing models have been created. A continuing difficulty, however, has been the creation of such models that can operate in or close to real time and preserve recognition accuracy.
As is well known in the art, one of the tools of signal recognition technology comprises the “hidden Markov model” (HMM). The HMM is a probabilistic pattern-matching approach which models a time-sequence of speech patterns as the result of random processes. The model is said to be ‘hidden’ because the initial state sequence that yields a given sequence of patterns cannot be determined. For example it is well known in the art that, the HMM is used in Carnegie Mellon's Sphinx-II system, a statistical modeling package for speech recognition.
In addition to recognizing a sequence of sounds as comprising a particular signature, which can be approached as a statistical problem, an interpretation of that sequence must also be made. This interpretation is known as decoding. This interpretation comprises searching for the most likely sequence of sounds given the input signal. One of the well-known methods known in the art is Viterbi decoding using a beam search, which is a dynamic programming algorithm that searches the state space for the most likely state sequence that accounts for the input signal. The state space is constructed by creating HMM models from the constituent HMM models, and the beam search is applied to limit the resulting large state space by eliminating less likely states. The Viterbi method is a time-synchronous search that processes the input signal one frame at a time and at a particular rate, typically 100 frames/sec.
The motion and formation of tornadoes have been difficult to predict, since they are highly localized, form quickly, and move erratically and very quickly. Although acoustic monitoring is widely utilized in weather tracking, there is need to improve the range, accuracy and processing time for determining the presence, location and movement of severe weather systems such as tornadoes.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide an improved signal recognition system for recognizing an approach or presence of an extreme weather system from an audio signal.
It is a further object to provide such a method for recognizing a tornado.
It is another object to provide such a method of building a set of software-based models for use in tornado recognition.
An additional object is to provide a system and method for generating a frequency-domain software system for use in signal processing applications.
A further object is to provide a system for providing early warning of a dangerous weather system such as a tornado.
These objects and others are attained by the present invention, an improved tornado recognition system and associated methods. A particular problem to be addressed herein is that people often ignore tornado warnings because sensors for tornadic activity are regional and are not deterministic. One aspect of the invention is a method and system for converting a sound signal containing a weather-indicating component and a noise component into a probability of a presence of an extreme weather system. The method comprises the steps of transforming the sound signal from a time domain into a frequency domain. Next the transformed signal is compared with a set of models of sound signals representative of a presence of the extreme weather system.
Next a determination is made from the comparing step of the probability of the presence of the extreme weather system. Finally, the determined probability is output, which enables a user to take an appropriate action.
In a particular embodiment of the invention for locating and tracking a tornado, a plurality of sound signals are collected in spaced-apart relation from each other to enable a triangulation of the signals if a tornadic acoustic signature is detected to a predetermined probability.
Another aspect of the invention is a method for determining a probability of a presence of an extreme weather system by collecting a sound signal that includes a tornado-indicating component and a noise component from an outdoor location. The sound signal is transformed from a time domain into a frequency domain. Next a training signal that comprises a weather-indicating component is transformed from a time domain to a frequency domain. Mel-banding is applied to the training signal. The mel-banded training signal is multiplied by a series of harmonically related cosine functions to obtain mel frequency cepstral coefficients. Hidden Markov models are built from the coefficients. Next the transformed sound signal is compared with at least one hidden Markov model representative of a presence of the extreme weather system. The probability for the presence of the extreme weather is determined. Finally the probability of an existence of a tornado is output to a location remote from the outdoor location.
In yet another aspect of the present invention the sound signal is collected from at least three outdoor locations in spaced relation from each other. The signal is transformed from a time domain into a frequency domain. A training signal comprising a weather-indicating component is transformed from a time domain to a frequency domain. Mel banding is performed on the training signal. The mel-banded training signal is multiplied by a series of harmonically related cosine functions to obtain mel frequency cepstral coefficients. Hidden Markov models are built from the coefficients. The transformed sound signal is compared with at least one hidden Markov model representative of a presence of the extreme weather system. A probability for the presence of the extreme weather is determined. Finally, the probability representative of a determination is output.
In yet a further aspect the method of building a set of models for determining a presence of a tornado from an outdoor weather sound signal comprises transforming a plurality of training sound signals from a time domain to a frequency domain, performing mel banding to the training signal, multiplying the mel-banded training signal by a series of harmonically related cosine functions to obtain mel frequency cepstral coefficients; and building hidden Markov models from the coefficients wherein building hidden Markov models comprises building five-state models.
In yet another embodiment of the present invention for converting a sound signal containing a weather-indicating component and noise into a probability of a presence of an extreme weather condition comprises a means for collecting the sound signal. The collecting means is positioned in an outdoor location wherein the collecting means comprises plurality of collecting means positioned in spaced relation from each other. A means for transforming the sound signal from a time domain into a frequency domain is provided. A means for transmitting the sound signal from the collecting means to the transforming means is also included. The embodiment includes a means for comparing the transformed sound signal with a set of models of all possible sound signals of the extreme weather condition. The comparing means receives the transformed sound signal from the transforming means and outputs a comparing signal. Next, a means for making a determination of a probability of a presence of the extreme weather condition by searching a set of control data models to match at least one data model with the transformed sound signal is included. The probability determination means receives the comparing signal and generates a probability signal. Finally, a
Baumgardner Jack W.
Cope R. Bradley
Abolafia Eliot
Craig Dwin M.
Jones Hugh
The United States of America as represented by the Secretary of
LandOfFree
Tornado recognition system and associated methods does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Tornado recognition system and associated methods, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Tornado recognition system and associated methods will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3314732