Data processing: measuring – calibrating – or testing – Measurement system – Measured signal processing
Reexamination Certificate
2000-06-02
2002-08-27
Shah, Kamini (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system
Measured signal processing
C073S589000, C367S125000
Reexamination Certificate
active
06442510
ABSTRACT:
DESCRIPTION
1. State of the Art
The invention relates to a method and apparatus to segment-wise detect and bring into coincidence signal waveforms which can be converted into monotone and continuous trajectories, in particular for purposes of real-time pattern recognition, of position finding and determination of transit-time differentials. Said signals are sample in time-discrete manner at the input side, are detected by correlation and thereupon the signal transit-times of different pairs of signal combinations and of different signal waveforms will then be determined.
The invention also relates to apparatus implementing such a method.
A procedure of the above kind used in signal detection is based on conventional template matching wherein the input signal is correlated with a pre-stored reference signal. By means of time-discrete sampling, the signal is rendered as a sequence of input vectors which form a matrix within which the signal is determined by the sum of all matrix pixel positions. In this matrix, correlation is determined by counting the coincident pixel positions between the input and the reference signal. This procedure is modified in time-sequence template matching in that an input vector is correlated with the search vector in each clock step and the partial results are stored in a memory and are summed after n clock steps into a total correlation result.
Time-sequence template matching is expanded into a parallel multi-template procedure with n search patterns by replicating the correlation unit n-fold and by mounting n correlation units next to each other. The input vectors pass through the correlation units and the correlation of an input vector with the comparison vector is carried out in each correlation unit.
A translationally invariant parallel, multi-template matching procedure is known from “Lazzaro, J. and Mead, C, ‘A silicon model of auditory localization’ (1989), Neural computation, 1, pp 47-57”. Therein the transit-time differential of acoustic signals is determined in two coupled functional units, namely an analog silicon cochlea which simulates the properties of a cochlea in biological systems and a subsequent correlation unit. At the input side, the silicon cochlea receives signals which drive the silicon cochlea into an energized state. Depending on the drive signal, sensors mounted equidistantly along the cochlea are actuated which relay the standard pulses to the subsequent correlation unit. This apparatus operates pairwise and the standard pulses received from the right or left cochlea pass through an antiparallel correlation unit consisting of two sets of antiparallel delay lines, each pair of delay lines being combined at the same sensor position. AND gates are mounted equidistantly along the delay lines and sample the signal status of the delay lines and, in the event of coincidence caused by standard pulses simultaneously incident on the AND gate, generate a coincidence signal at the gate output. The correlation unit determines the coincidences of the standard pulses at each pair of lines and the number of coincidences along the same positioning of the AND gates in the direction of the delay lines is counted. The output is a vector of signal-transit differentials per clock step. Accordingly the known procedure is based on simulating a cochlea and determining coincidences along analog delay lines.
Furthermore U.S. Pat. No. 5,417,113 discloses a procedure allowing to localize sound sources by means of test data from several silicon cochlea. The microphone signals are fed pair-wise to several silicon cochlea carrying out a 2-D correlation along predetermined delay lines. The detected coincidences are fed from the particular silicon cochlea to a subsequent 3D analyzer which, upon comparing the characteristic signatures of the 2D output data, results in improved localization of the sound source.
2. Drawbacks of the State of the Art
The conventional template-matching procedures incur the drawback that they fail to be invariant in the presence of spatial shifts. Only a signal sequence present at the same time of synchronization at which the correlation between input and comparison vector is started will result in maximum correlation. If the correlation between the input and the comparison vector is shifted by one or more sequences, the image and search patterns no longer coincide.
The known translationally invariant, parallel, multi-template matching procedures used to determine the transit-time differential of acoustic signals incur the drawback of the complex simulation of a biological cochlea, in particular in tuning the response of the cochlea to different energizing signals by varying the relevant parameters describing the cochlea.
Another drawback resulting from the circuitry of the heretofore designed analog configurations is the adjustment, subject to manufacturing tolerances, of the delay speeds in the analog delay lines subject.
The heretofore predominantly employed digital signal processors (DSP's) incur a drawback in that they do not properly match the present problem's solution contained in the method and apparatus of the invention. A DSP is incapable of solving said problem of a defined magnitude within an adequately short time, and in many cases such inability may entail problems.
A further general drawback of the DSP's prevailingly used to-date for signal processing and pattern recognition is that the procedure and the apparatus poorly match the above solution to the problem addressed herein. A DSP is incapable of solving such a problem of a defined magnitude at the same time as the method and apparatus herein disclosed. The DSP architecture cannot process the signal flows in the manner of the method and apparatus of the invention.
Another general drawback of the DSP's is the central control unit and the scarcity of computers able to process the data flows in the manner of the method and apparatus of the present invention. As regards digital signal processors, instructions are stored in the computers, operands are retrieved, and results are filed in registers. Because of its fixed CPU architecture, a digital signal processor is unable to simulate the apparatus'signal-flow architecture at which the signal flows pass in time though said apparatus.
OBJECTIVE OF THE INVENTION
The objective of the invention is to create a method and apparatus to segment-wise detect and bring into coincidence signal waveforms which can be converted into monotone and continuous trajectories, in particular to recognize patterns in real time, to localize and to monitor optical and acoustic signals, and to determine transit differentials with increased accuracy of measurement by making it possible to program the key waveforms and the resolution.
Another objective of the invention is to accelerate calculations in order to include several domains of application of real-time pattern recognition.
EXAMPLES
In the method of the invention, first pattern recognition is carried out and then the determination of the transit-time differential by detecting key waveforms, and the information is processed further in a subsequent multi-coincidence unit; shift-invariant, parallel multi-template matching is carried out during which key waveforms are correlated in parallel and an output per clock step, illustratively of a set of transit-time differentials, is created in the said subsequent multi-coincidence unit.
The method to segment-wise determine pattern recognition and transit-time differentials of signal waveforms which can be converted into monotone and continuous space-time trajectories, in particular for purposes of real-time pattern recognition, is characterized in that pre-programmed key signals are detected by signal sampling and processing the sampled data, further by subjecting to multiple coincidences combined signal pairs of different signal transit times and different waveforms. At the input side the signals are sampled in a sequence of input vectors, each input vector consecutively passing through a signal detection unit consisti
Clark & Brody
Shah Kamini
LandOfFree
Method and apparatus for determining transit-time... 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 apparatus for determining transit-time..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for determining transit-time... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2890110