Communications device responsive to spoken commands and methods of using same

• US 5,749,072 A
• Filed: 12/28/1995
• Issued: 05/05/1998
• Est. Priority Date: 06/03/1994
• Status: Expired due to Fees

• Current Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Levendel, Gil E., Wang, Shay-Ping Thomas, Mazurkiewicz, Theodore
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Dorvil, Richemond
• Application Number:
  US08/579,714
• Time in Patent Office: 859 Days
• Field of Search: 395/2.41, 395/2.79, 395/2.84, 395/2.62, 395/2.6, 395/2.64, 395/2.63, 395/2.65, 395/2.66, 395/2.49, 704/275, 704/245, 704/270, 704/253, 704/255, 704/251, 704/254, 704/256, 704/257, 704/239, 704/240
• US Class Current: 704/275
• CPC Class Codes: G06F 18/2453 non-linear, e.g. polynomial classifier
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 15/16 using artificial neural networks
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

A communications device (20) that is responsive to voice commands is provided. The communications device (20) can be a two-way radio, cellular telephone, PDA, or pager. The communications device (20) includes an interface (22) for allowing a user to access a communications channel according a control signal and a speech-recognition system (24) for producing the control signal in response to a voice command. Included in the speech recognition system (24) are a feature extractor (26) and one or more classifiers (28) utilizing polynomial discriminant functions.

Claims

1. A communications device, comprising:

an interface for allowing a user to access a communications channel according a control signal; and

a speech-recognition system for producing the control signal in response to a spoken command, the speech-recognition system including;

a feature extractor for extracting a plurality of features from the spoken command; and

a classifier for generating a discriminant signal according to a polynomial expansion having a form ##EQU4## wherein x_j represents the plurality of features, y represents the discriminant signal, w_i represents a coefficient, g_ji represents an exponent, and i, j, m and n are integers;

wherein the control signal is based on the discriminant signal.

2. The communications device of claim 1, wherein the polynomial expansion has a form ##EQU5## wherein a₀ represents a zero-order coefficient, b_i represents a first-order coefficient, and c_ij represents a second-order coefficient.

3. The communications device of claim 1, wherein the interface includes a device selected from a group consisting of:

a two-way radio, a telephone, a PDA, and a pager.

4. The communications device of claim 1, wherein the spoken command is a word selected from a group consisting of a digit between 0-9, "page", "send", and "help".

5. The communications device of claim 1, wherein the speech-recognition system further comprises:

a pre-processor, operatively associated with the feature extractor, for transforming an audio signal using signal processing techniques into a sequence of data vectors that represent the spoken command and from which the plurality of features are extracted.

6. The communications device of claim 1, wherein the plurality of features are selected from a group consisting of:

cepstral coefficients, first-order derivatives of cepstral coefficients, and word-level features.

7. A communications device, comprising:

a pre-processor for transforming an audio signal into a sequence of data vectors;

extraction means for extracting a plurality of feature frames from the sequence of data vectors;

a plurality of classifiers for generating a plurality of discriminant signals, each of the plurality of classifiers designating a different spoken command and generating a discriminant signal according to a polynomial expansion having a form ##EQU6## wherein x_j represents a feature frame, y represents the discriminant signal, w_i represents a coefficient, g_ji represents an exponent, and i, j, m and n are integers;

an accumulator for generating a plurality of accumulated discriminant signals, the accumulator generating each of the plurality of accumulated discriminant signals by summing ones of the plurality of discriminant signals produced by a respective one of the plurality of classifiers;

a selector for selecting a largest accumulated discriminant signal from the plurality of accumulated discriminant signals; and

a two-way audio interface for transmitting and receiving data across a communications channel according a control signal, the control signal being a function of the largest accumulated discriminant signal.

8. The communications device of claim 7, wherein the extraction means includes:

a feature extractor for extracting a sequence of feature frames from the sequence of data vectors; and

a speech activity detector for selecting from the sequence of feature frames the plurality of feature frames representing a spoken command.

9. The communications device of claim 7, wherein the extraction means includes:

a speech activity detector for selecting from the sequence of data vectors a vector sub-sequence representing a spoken command; and

a feature extractor for extracting a plurality of feature frames from the vector sub-sequence.

10. The communications device of claim 7, wherein the polynomial expansion has a form ##EQU7## wherein a₀ represents a zero-order coefficient, b_i represents a first-order coefficient, and c_ij represents a second-order coefficient.

11. The communications device of claim 7, wherein the two-way audio interface includes a device selected from a group consisting of:

a two-way radio, a telephone, a PDA, and a pager.

12. The communications device of claim 7, wherein the audio signal represents a spoken command selected from a group consisting of a digit between 0-9, "page", "send", and "help".

13. The communications device of claim 7, wherein each of the plurality of feature frames includes a plurality of features selected from a group consisting of:

cepstral coefficients, first-order derivatives of cepstral coefficients, and word-level features.

14. A two-way handheld communications device, comprising:

a microphone for generating an audio signal;

an A/D converter for digitizing the audio signal to produce a digitized audio signal;

a pre-processor for transforming the digitized audio signal into a sequence of data vectors;

a speech activity detector for producing a vector sub-sequence representing a spoken command, the speech activity detector continuously receiving the sequence of data vectors and including in the vector sub-sequence those of the sequence of data vectors having an energy-level that exceeds a background noise threshold;

a feature extractor for extracting a sequence of feature frames from the vector sub-sequence;

a plurality of classifiers for generating a plurality of discriminant signals, each of the plurality of classifiers designating a different spoken command and generating a discriminant signal according to a polynomial expansion having a form ##EQU8## wherein x_j represents a feature frame, y represents the discriminant signal, w_i represents a coefficient, g_ji represents an exponent, and i, j, m and n are integers;

a plurality of accumulators for generating a plurality of accumulated discriminant signals, each of the accumulators summing ones of the plurality of discriminant signals produced by a respective one of the plurality of classifiers;

a selector for selecting a largest accumulated discriminant signal from the plurality of accumulated discriminant signals; and

a two-way audio interface for transmitting and receiving data across a radio channel according a control signal, the control signal being a function of the largest accumulated discriminant signal.

15. The two-way handheld communications device of claim 14, wherein the polynomial expansion has a form ##EQU9## wherein a₀ represents a zero-order coefficient, b_i represents a first-order coefficient, and c_ij represents a second-order coefficient.

16. The two-way handheld communications device of claim 14, wherein the two-way audio interface includes a device selected from a group consisting of:

a two-way radio, a telephone, a PDA, and a pager.

17. The two-way handheld communications device of claim 14, wherein the spoken command is a word selected from a group consisting of a digit between 0-9, "page", "send", and "help".

18. The two-way handheld communications device of claim 14, wherein the speech activity detector detects boundaries of the spoken command by determining energy-level transitions across the background noise threshold.

19. The two-way handheld communications device of claim 18, wherein the speech activity detector associates an end-of-word boundary with a negative energy-level transition if the energy-level remains below the background noise threshold during a subsequent predetermined interval.

20. A method for controlling access to a communications channel, comprising the following steps:

receiving a spoken command;

extracting a plurality of features from the spoken command;

generating a discriminant signal based on a polynomial expansion having a form ##EQU10## wherein x_j represents the plurality of features, y represents the discriminant signal, w_i represents a coefficient, g_ji represents an exponent, and i, j, m and n are integers; and

accessing the communications channel according the discriminant signal.

21. The method of claim 20, wherein the step of generating includes the following sub-step:

basing the discriminant signal on a second-order polynomial expansion having a form ##EQU11## wherein a₀ represents a zero-order coefficient, b_i represents a first-order coefficient, and c_ij represents a second-order coefficient.

22. The method of claim 20, further comprising the following step:

selecting the spoken command from a group consisting of a digit between 0-9, "page", "send", and "help".

23. The method of claim 20, further comprising the step of:

transforming an audio signal using signal processing techniques into a sequence of data vectors that represent the spoken command and from which the plurality of features are extracted.

24. The method of claim 20, wherein the step of extracting includes the following sub-step:

generating the plurality of features selected from a group consisting of;

cepstral coefficients, first-order derivatives of cepstral coefficients, and word-level features.