Apparatus and method for continuous speech recognition on a PCMCIA card

• US 6,804,642 B1
• Filed: 06/24/1997
• Issued: 10/12/2004
• Est. Priority Date: 06/24/1997
• Status: Expired due to Fees

• Current Assignee: Exelis Incorporated (L3Harris Technologies, Inc.)
• Original Assignee: ITT Manufacturing Enterprises, Inc. (ITT, Inc.)
• Inventors: Naylor, Joe A.
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Opsasnick, Michael N.
• Application Number:
  US08/881,741
• Time in Patent Office: 2,667 Days
• Field of Search: 704/246, 704/248, 704/250
• US Class Current: 704/231
• CPC Class Codes: G10L 15/285 Memory allocation or algorithm optimisation to reduce hardware requirements

Abstract

An apparatus and method for automatic continuous-speech recognition is disclosed. The recognition function for a speaker-dependent continuous-speech recognition (CSR) system is performed on a (PCMCIA) card, which includes a DSP, Codec, and memory. All of the processing performed by the CPU, DSP, and special DTW chip in prior art systems are performed in the single DSP on the PCMCIA card. Recognition results computed on the card are uploaded to the host computer application (for command interpretation and processing), via the PCMCIA interface. The host computer may comprise a personal computer, making this invention particularly well-suited for use in military or other such operations which require light-weight portable computers.

Claims

1. In a continuous-speech recognition system for the recognition of words spoken by an individual, including the steps of receiving input speech data, digitizing the speech data, and analyzing the speech data by a speech recognition module which compares the input speech data to previously obtained templates and generates respective match scores therefrom, the improvement wherein:

the continuous-speech recognition system is implemented on a system consisting of a personal computer host with a peripheral PCMCIA card; and

,

PCMCIA executable modules are stored on a host disk associated with said personal computer host and are downloaded to said PCMCIA card at run time so that the modules may be easily modified at any time, and wherein PCMCIA memory is accessible (read and write) by the personal computer host.

2. The continuous-speech recognition system of claim 1, wherein the PCMCIA card runs speech recognition software and provides results to the personal computer host in the form of a string of words, and wherein the personal computer host converts the word strings to commands for controlling applications through socket or server connections.

3. The continuous-speech recognition system of claim 1, wherein the personal computer host includes system hardware with specifications for interfacing with the PCMCIA and meeting requirements for the PCMCIA card, including a digital signal processor, memory, and codec.

4. The continuous-speech recognition system of claim 3, wherein the gain and sample rate functions of the PCMCIA codec are software controllable.

5. The continuous-speech recognition system of claim 3, further comprising programs and files for creating syntax files and training word models for an individual for specific continuous-speech recognition tasks.

6. The continuous-speech recognition system of claim 3, further comprising programs for performing recognition tasks with a trained speaker'"'"'s word models, including software hooks for translating word strings into commands for socket-type control of application programs.

7. The continuous-speech recognition system of claim 1, wherein the PCMCIA card has a microphone interface disposed on its exterior receptacle which allows live speech to be digitized directly into PCMCIA memory.

8. The continuous-speech recognition system of claim 1, wherein said personal computer host is portable and lightweight.

9. In a continuous-speech recognition system, including the steps of receiving input training speech data during training procedures, digitizing the training speech data, converting the training speech data to filter bank parameters, storing the parameters of the training speech data as word templates, receiving input test speech data, digitizing the test speech data, converting the test speech data to filter bank parameters, and analyzing the test speech data by a speech recognition module which applies speech recognition algorithms to compare the test speech data to the templates and generate respective match scores therefrom, the improvement wherein:

the continuous-speech recognition system is implemented on a system consisting of a host computer and a speech recognition processor;

the host computer comprises a personal computer;

the speech recognition processor comprises a PCMCIA card that receives the input speech data, digitizes the speech data, and performs the match-scoring between the test speech data and the stored word templates;

the PCMCIA card is peripheral to the host computer; and

,

an executable operative to direct the functioning of said PCMCIA card is downloaded from said host computer to said PCMCIA card at runtime.

10. The continuous speech recognition system of claim 9, wherein said PCMCIA card can be reprogrammed without the need to burn-in new firmware.

11. The continuous-speech recognition system of claim 10, wherein the host computer includes hardware having drivers for controlling the PCMCIA card and accessing its memory.

12. The continuous-speech recognition system of claim 11, wherein the host computer includes software for performing syntax generation, template training, and a command interpreter for translating the recognizer output stream into commands for socket-type control of tasks by voice command.

13. The continuous-speech recognition system of claim 10, further comprising a microphone on the exterior of the PCMCIA card for receiving the input speech data.

14. The continuous-speech recognition system of claim 10, wherein the PCMCIA card runs speech recognition software and provides results to the personal computer host in the form of a string of words, and wherein the personal computer host converts the word strings to commands for controlling applications through socket or server connections.

15. The continuous-speech recognition system of claim 14, wherein the personal computer host includes system hardware with specifications for interfacing with the PCMCIA and meeting requirements for the PCMCIA card, including a digital signal processor, memory, and codec.