Parallel signal processing system and method

• US 9,832,543 B1
• Filed: 06/16/2014
• Issued: 11/28/2017
• Est. Priority Date: 09/29/2008
• Status: Active Grant

• Current Assignee: Patent Armory Inc.
• Original Assignee: Calltrol Corporation
• Inventors: Wu, Wai
• Primary Examiner(s): Nguyen, Duc
• Assistant Examiner(s): Mohammed, Assad
• Application Number:
  US14/305,432
• Time in Patent Office: 1,261 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 signal streams within a plurality of channels, comprising:

(a) providing a plurality of timeslices of the signal streams within the plurality of channels, each timeslice representing a time-continuous portion of digitized data of a respective signal stream within a respective channel;

(b) loading the plurality of timeslices into a memory accessible by at least one multiprocessor, each respective multiprocessor having a plurality of processing cores, each of the plurality of processing cores being controlled in accordance with a common instruction sequence for the respective multiprocessor, such that the plurality of processing cores of a respective multiprocessor synchronously and concurrently execute the same processing instruction on data of a different signal stream;

(c) providing the common instruction sequence for the multiprocessor adapted to process each timeslice to produce at least one of processed data and at least one data processing result for each respective timeslice;

(d) executing the common instruction sequence on the at least one multiprocessor, to produce at least one of processed data and at least one data processing result for each respective timeslice, comprising at least one of a time-frequency domain transform and a wavelet transform of the time-continuous portion of digitized data of each respective timeslice in parallel; and

(e) storing the at least one of processed data and at least one data processing result for each respective timeslice in a memory.

2. The method according to claim 1, wherein the signal streams each comprise telephone signals, and the common instruction sequence comprises a call progress tone analysis algorithm.

3. The method according to claim 1, wherein the common instruction sequence executes substantially without data-dependent conditional execution branch instructions.

4. The method according to claim 1, wherein the common instruction sequence executes substantially without interaction between respective channels.

5. The method according to claim 1, wherein the at least one multiprocessor is provided as a coprocessor to a general purpose processor, further comprising transferring the stored at least one of processed data and at least one data processing result for each respective timeslice from a first memory device associated with the at least one multiprocessor to a second memory device associated with the general purpose processor.

6. The method according to claim 1, wherein each signal stream comprises a telephone communication stream, and the common instruction sequence controls each respective processing core to concurrently execute a call progress tone detection algorithm,

further comprising responding to a detected call progress tone represented in at least one telephone communication stream.

7. The method according to claim 1, wherein the common instruction sequence controls each respective processing core to concurrently execute a time-frequency domain transform.

8. The method according to claim 1, wherein the common instruction sequence controls each respective processing core to concurrently execute a wavelet transform.

9. The method according to claim 1, wherein the signal streams comprise human voice, and the data processing result is employed to perform speech recognition for a respective channel.

10. The method according to claim 1, wherein the common instruction sequence comprises a Goertzel filter algorithm.

11. The method according to claim 1, wherein the multiprocessor processes sequential timeslices of a plurality of audio data channels with real-time throughput.

12. The method according to claim 1, wherein the multiprocessor is part of a graphic processing unit, further comprising using the stored processing result to control a telephone switching system.

13. The method according to claim 1, wherein the common instruction sequence defines a Fourier transform.

14. The method according to claim 1, wherein the stored at least one of processed data and at least one data processing result for each respective time-slice in a memory is analyzed to determine an existence of an predetermined in-band signal.

15. The method according to claim 1, wherein the at least one of processed data and at least one data processing result for each respective timeslice represents a joint processing result representing a contribution from each of a plurality of independent data streams, wherein the joint processing result is produced in real time at a fixed maximum latency after the a continuous portion of digitized data is stored in the memory.

16. The method according to claim 15, wherein the plurality of independent data streams comprise a plurality of telephone communication channels, wherein the common instruction sequence defines an algorithm adapted to process each timeslice to identify an in-band signaling tone of a respective telephone communication channel.

17. The method according to claim 16, wherein the instruction sequence defines a Goertzel filter algorithm to identify the in-band signaling tone, further comprising receiving the identification of the in-band signaling tone by a host general purpose programmable processor and performing a communication channel control function by the general purpose programmable processor in dependence on the identification.

18. A method for processing a set of parallel signal streams, comprising:

(a) receiving a series of timeslices over time for each respective parallel signal stream, each timeslice representing a time-continuous portion of digitized data of a respective signal stream loaded into a memory accessible by a multiprocessor having a plurality of parallel processing cores, the plurality of parallel processing cores being under control of a common instruction sequence, such that the plurality of parallel processing synchronously and concurrently execute the same processing instruction on a respective timeslice of a different signal stream; and

(b) executing the common instruction sequence on the multiprocessor, to produce a respective processing result for each of the parallel signal streams, comprising convolution processing and at least one of a time-frequency domain transform and a wavelet transform of the time-continuous portion of digitized data of each respective timeslice.

19. A parallel processing method for processing a plurality of audio channels comprising speech in parallel, comprising:

receiving the plurality of audio channels comprising speech in parallel;

loading a current set of parallel timeslices from the plurality of audio channels comprising speech into a memory array accessible by a multiprocessor system, the multiprocessor system having a plurality of processing cores together controlled in parallel by a common instruction sequence;

processing the current set of parallel timeslices from the plurality of audio channels comprising speech with the multiprocessor system, according to the common instruction sequence such that each processing core of the multiprocessor is synchronized to concurrently execute the same instruction on data representing a timeslice of a respective audio channel, to at least one of characterize the speech of a respective audio channel, characterize an automate in-band signal of a respective audio channel, and characterize a respective audio channel, to produce a set of current characterizations for each of the set of parallel timeslices stored in a memory;

loading a subsequent set of parallel timeslices from the plurality of audio channels comprising speech into the same memory locations of the memory array as previously occupied by the current set of parallel timeslices; and

processing the subsequent set of parallel timeslices from the plurality of audio channels comprising speech according to an algorithm comprising at least one of a time-frequency domain transform and a wavelet transform with the multiprocessor system, according to the common instruction sequence such that each processing core of the multiprocessor is synchronized to concurrently execute the same instruction on data representing a timeslice of a respective audio channel, to at least one of characterize the speech of a respective audio channel, characterize an automate in-band signal of a respective audio channel, characterize the respective audio channel, and to determine a change in a respective prior characterization associated with a respective audio channel, to produce a set of subsequent characterizations for each of the set of parallel timeslices stored in the memory.