Parallel signal processing system and method

• US 10,524,024 B1
• Filed: 11/27/2017
• Issued: 12/31/2019
• Est. Priority Date: 09/29/2008
• Status: Expired due to Fees

• Current Assignee: Patent Armory Inc.
• Original Assignee: Calltrol Corporation
• Inventors: Wu, Wai
• Primary Examiner(s): Nguyen, Duc
• Assistant Examiner(s): Mohammed, Assad
• Application Number:
  US15/823,430
• Time in Patent Office: 764 Days
• CPC Class Codes: H04M 11/007 with remote control systems remote control in general G08C
• CPC Class Codes: H04M 11/066 Telephone sets adapted for data transmision
• CPC Class Codes: H04M 5/00 Manual exchanges substation equipment in general H04M1/00
• CPC Class Codes: H04Q 3/545 using a stored programme

Abstract

A system and method for processing a plurality of channels, for example audio channels, in parallel is provided. For example, a plurality of telephony channels are processed in order to detect and respond to call progress tones. The channels may be processed according to a common transform algorithm. Advantageously, a massively parallel architecture is employed, in which operations on many channels are synchronized, to achieve a high efficiency parallel processing environment. The parallel processor may be situated on a data bus, separate from a main general purpose processor, or integrated with the processor in a common board or integrated device. All, or a portion of a speech processing algorithm may also be performed in a massively parallel manner.

Claims

1. A method for processing streams of information, comprising:

receiving a plurality of streams of information, wherein each respective stream of information represents a distinct voice communication channel carrying a different voice communication;

storing data representing the plurality of streams of information in a memory;

accessing the memory with a graphic processing unit comprising an instruction decoder controlling a plurality of arithmetic logic units;

concurrently processing the data representing the plurality of streams of information in parallel in accordance with a common sequence of instructions, wherein the common sequence of instructions performs at least a portion of a speech recognition process on each respective distinct voice communication channel; and

storing the processed data in the memory.

2. The method according to claim 1, wherein the common set of instructions executes substantially without data-dependent conditional execution branch instructions.

3. The method according to claim 1, wherein the common set of instructions executes substantially without interaction between respective streams of information.

4. The method according to claim 1, wherein the graphics processing unit is provided as a coprocessor to a general purpose processor, further comprising the step of transferring the data representing the plurality of streams of information from general purpose processor memory to graphics processing unit memory before said accessing.

5. The method according to claim 1, wherein the common set of instructions comprises a call progress tone detection algorithm, further comprising issuing a signal in response to detection of a respective type of call progress tone represented in a respective stream of information.

6. The method according to claim 1, wherein the common set of instructions performs at least one speech recognition primitive.

7. The method according to claim 1, wherein the common set of instructions performs echo cancellation.

8. The method according to claim 1, wherein the common set of instructions performs a data domain transformation.

9. The method according to claim 1, wherein the common set of instructions implements a transversal filter.

10. The method according to claim 1, wherein each respective stream of information has a respective data rate, and wherein the graphics processing unit processes the data representing each stream of information at the respective data rate to provide real-time processing.

11. The method according to claim 1, further comprising producing a decision based on each respective stream of information, and controlling a communication associated with the stream of information based on the respective decision.

19. The method according to claim 1, wherein each of the plurality of streams of information comprises a separate human speech communication, and the common sequence of instructions performs at least a portion of the speech recognition process performs the same portion of the speech recognition on each of the separate human speech communications.

12. A method for processing a plurality of streams of speech information, comprising:

receiving the plurality of streams of speech information;

concurrently processing each respective stream of speech information in parallel, with a graphic processing unit comprising an instruction decoder controlling a plurality of arithmetic logic units in accordance with a common set of instructions, to concurrently perform at least a portion of a speech recognition process on a respective different stream of speech information of the plurality of streams of speech information across the plurality of arithmetic logic units;

storing the processed plurality of streams of speech information in a memory; and

recognizing speech within at least one stream of the plurality of streams of speech information.

13. The method according to claim 12, wherein the plurality of streams of speech information are stored in the memory as a sequence of time slices, each time slice representing a time-continuous portion of data, which are sequentially concurrently processed by the graphic processing unit across the plurality of arithmetic logic units.

14. The method according to claim 12, further comprising performing echo processing on the plurality of streams of speech information, in parallel in real time by the graphic processing unit.

15. The method according to claim 12, wherein the common set of instructions implements a decision-making function.

16. The method according to claim 12, further comprising performing transversal filter function in parallel on the plurality of streams of speech information.

20. The method according to claim 12, wherein each respective stream of speech information comprises a separate human speech communication, and the common set of instructions concurrently performs the same portion of the speech recognition process on each respective each respective stream of speech information.

17. A non-transitory computer readable medium storing therein non-transitory instructions for controlling a graphic processing unit comprising an instruction decoder controlling a plurality of arithmetic logic units in parallel, comprising:

instructions for accessing a memory associated with the graphic processing unit;

common instructions, for decoding by the instruction decoder, for concurrently processing data representing a plurality of streams of information in parallel on the plurality of arithmetic logic units, wherein the plurality of streams of information comprise distinct speech communications, and the common instructions comprise common instructions for concurrently performing at least a portion of a speech recognition process on each respective distinct speech communication in parallel; and

instructions for storing the processed data from the plurality of arithmetic logic units in the memory.

18. The non-transitory computer readable medium according to claim 17, wherein the processed data comprises a joint processing result representing a contribution from each of a plurality of streams of information.