Method for signal processing

• US 4,672,667 A
• Filed: 06/02/1983
• Issued: 06/09/1987
• Est. Priority Date: 06/02/1983
• Status: Expired due to Fees

• Current Assignee: Scott Instruments Company Denton TX
• Original Assignee: Scott Instruments Company Denton TX
• Inventors: Newell, John M., Scott, Brian L.
• Primary Examiner(s): KEMENY, EMANUEL
• Application Number:
  US06/500,548
• Time in Patent Office: 1,468 Days
• Field of Search: 381/29-53, 364/513, 364/513.5, 364/604, 364/728
• US Class Current: 704/231
• CPC Class Codes: G10L 15/24 Speech recognition using non-acoustical features

Abstract

A method and apparatus for processing a signal to extract selected information therefrom is provided. According to the method, an input signal is converted into a sequence of data samples, and these data samples are applied sequentially through a convolver having first and second sections, the output of the first section forming the input of the second section. A data sample in the second convolver section is then compared to each data sample in the first convolver section to produce an output signal of a plurality of data points, each of the data points representative of a midpoint position within the convolver between a pair of compared data samples. This comparison step is repeated for each data sample in the second convolver section and the output signals form a histogram from which the selected information is extracted. This method may be advantageously utilized in a speech recognition process for extracting various features from an input speech signal.

Claims

1. A method for processing a signal to extract selected information therefrom, comprising the steps of:

(a) converting said signal into first and second sequences of data samples, the first sequence including samples 1 . . . (n/2) and the second sequence including samples (n/2)+1 . . . n and the end of said first sequence forming the beginning of said second sequence;

(b) translating the first sequence in a first direction and the second sequence in a second direction opposite of the first direction such that samples 1 . . . (n/2) of the first sequence positionly overlie samples n . . . (n/2)+1 of the second sequence;

(c) comparing a position of a data sample in said second sequence to the position of each data sample in said first sequence to produce a plurality of data points, each of said data points representative of a midpoint position between the compared data samples;

(d) repeating step (c) for the position of each data sample in said second sequence; and

(e) summing the data points at each midpoint position to form a histogram.

2. The method for processing as described in claim 1 wherein the step of converting comprises converting said signal from an analog form to a digital form.

3. The method for processing as described in claim 1 further including the step of sampling said sequence of data samples to produce a sequence of peak data samples.

4. The method for processing as described in claim 1 wherein symmetrical patterns are represented in said histogram when said signal is periodic.

5. The method for processing as described in claim 4 wherein said symmetrical patterns are separated by a distance equaling the period of the fundamental frequency of said periodic signal.

6. The method for processing as described in claim 1 wherein said midpoint position of the compared data samples is determined by calculating the distance between said compared data samples and dividing by two.

7. The method for processing as described in claim 1 wherein said data samples are compared only if the amplitudes of said data samples are within a predetermined range.

8. The method for processing as described in claim 7 wherein said data points have a value proportional to the amplitudes of said compared data samples.

9. A method for processing a speech signal to extract selected information therefrom, said selected information useful for speech recognition, comprising the steps of:

(a) converting said speech signal into first and second sequences of digital samples, the end of the first sequence forming the beginning of the second sequence;

(b) sampling said first and second sequences of data samples to produce first and second sequences of peak data samples, the first sequence including peak data samples 1 . . . (n/2) and the second sequence including peak data samples (n/2)+1 . . . n;

(c) translating the first sequence of peak data samples in a first direction and the second sequence of peak data samples in a second direction opposite of the first direction such that samples 1 . . . (n/2) of the first sequence positionly overlie samples n . . . (n/2)+1 of the second sequence;

(d) comparing a position of a data sample in said second sequence of peak data samples to each data sample in said first sequence of peak data samples to produce a plurality of data points, each of said data points representative of a midpoint position between the compared data samples;

(e) repeating step (d) for the position of each data sample in said second sequence of peak data samples;

(f) summing the data points at each midpoint position to form a histogram.

10. The method for processing a speech signal as described in claim 9 wherein said histogram has symmetrical patterns separated by a distance equaling the period of the fundamental frequency of said speech signal when said speech signal is periodic.

11. The method for processing a speech signal as described in claim 9 wherein the midpoint position is determined by calculating the distance between the compared data samples and dividing by two.

12. The method for processing a speech signal as described in claim 9 wherein said data samples are compared only when the amplitudes of said data samples are within a predetermined range.

13. The method for processing a speech signal as described in claim 12 wherein said data points have a value proportional to said amplitude of said compared data samples.

14. A method for processing a signal to extract selected information therefrom, comprising the steps of:

(a) converting said signal into first and second sequences of data samples, the first sequence including samples 1 . . . (n/2) and the second sequence including samples (n/2)+1 . . . n, each of the first and second sequences having a beginning and an end, the end of the second sequence forming the beginning of the first sequence;

(b) translating the first sequence of data samples in a first direction and the second sequence of data samples in a second direction opposite of the first direction such that samples 1 . . . (n/2) of the first sequence positionly overlie samples n . . . (n/2)+1 of the second sequence;

(c) comparing the positions of the data samples at the end of the first and second sequences, respectively, with the position of each data sample in said second and first sequences, respectively, to produce a pair of output signals each having a plurality of data points, each of said data points representative of a midpoint position between the compared data samples;

(d) addition a new data sample to said first sequence and shifting the data samples in the first and second sequences by one position;

(e) comparing the positions of the data samples at the beginning of the first and second sequences, respectively, with the position of each data sample in said second and first sequences, respectively, to produce a second pair of output signals;

(f) repeating steps (c)-(e) for additonal data samples; and

(g) summing the data points at each midpoint position generated in step (e) and subtracting the data points at each midpoint position generated in step (c) to form a histogram.

15. The method for processing as described in claim 14 wherein the step of converting comprises converting said signal from an analog form to a digital form.

16. The method for processing as described in claim 14 further including the steps of sampling said sequence of data samples to produce a sequence of peak data samples.

17. The method for processing as described in claim 14 wherein said histogram has symmetrical patterns when said signal is periodic.

18. The method for processing as described in claim 17 wherein said symmetrical patterns are separated by a distance equaling the period of the fundamental frequency of said periodic signal.

19. The method for processing as described in claim 14 wherein said midpoint position is determined by calculating the distance between said compared data samples and dividing by two.

20. The method for processing as described in claim 14 wherein said data samples are compared only when the amplitudes of said data samples are within a predetermined range.

21. The method for processing as described in claim 20 wherein said data points have a value proportional to the amplitude of said compared data samples.