Acoustic sound signature detection based on sparse features

• US 9,785,706 B2
• Filed: 08/28/2013
• Issued: 10/10/2017
• Est. Priority Date: 08/28/2013
• Status: Active Grant

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Tan, Bozhao
• Primary Examiner(s): Goins, Davetta W
• Assistant Examiner(s): Ganmavo, Kuassi
• Application Number:
  US14/013,020
• Publication Number:
  US 20150063575A1
• Time in Patent Office: 1,504 Days
• Field of Search: 381 56, 381321, 381 941- 948, 704243, 7042562, 7042563, 704232, 704246, 704E15004
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• CPC Class Codes: G06F 16/683 using metadata automatically derived from the content
• CPC Class Codes: G06Q 40/08 Insurance
• CPC Class Codes: G07C 5/085 using electronic data carriers
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 25/00 Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 muting semiconductor-based amplifiers when some special characteristics of a signal are sensed by a speech detector, e.g. sensing when no signal is present, H03G3/34
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals
• CPC Class Codes: H04R 29/00 Monitoring arrangements; Testing arrangements for hearing aids H04R25/30; detection of loudspeaker connection H04R5/04; sound-field adaptation dependent on speaker detection H04S7/308
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A low power sound recognition sensor is configured to receive an analog signal that may contain a signature sound. Sparse sound parameter information is extracted from the analog signal. The extracted sparse sound parameter information is processed using a sound signature database stored in the sound recognition sensor to identify sounds or speech contained in the analog signal, wherein the sound signature database comprises a plurality of sound signatures each representing an entire word or multiword phrase.

Claims

1. A process of operating a sound recognition sensor comprising:

(a) receiving analog sound signals in signal trigger logic that is maintained in a powered up active mode while maintaining other circuitry of the sensor in a low power mode;

(b) powering up normally powered down parallel analog channels of the sensor in response to a first trigger signal indicating the signal trigger logic receiving analog sound signals greater than a reference;

(c) extracting a first frame based sparse feature from the received analog sound signals in one analog channel;

(d) extracting a second frame based sparse feature, different from the first frame based sparse feature, from the received analog sound signals in a another analog channel simultaneously with the extracting a first frame based sparse feature;

(e) producing an event trigger signal from event trigger logic in response to at least one of the extracted features from the channels indicating a starting point;

(f) powering up normally powered down classifier circuitry in response to the event trigger signal; and

(g) performing sound recognition on the received sound signals in the classifier circuitry based on the received analog sound signals.

2. The process of claim 1 in which the extracting the first and second frame based sparse features includes extracting sparse features from which speech content cannot be reconstructed.

3. The process of claim 1 in which the extracting the sparse sound features includes extracting one of zero crossing information, total energy values, band pass energy values, high pass energy values, or low pass energy values in each analog channel.

4. The process of claim 1 in which the performing sound recognition includes recognizing a portion of a spoken command of one second in duration.

5. A sound recognition sensor comprising:

(a) a preamplifier having an input receiving analog sound signals and an output providing amplified analog sound signals;

(b) signal trigger logic having an input coupled to the amplified analog sound signals and a sound detected output, the signal trigger logic being maintained in a powered mode, the signal trigger logic producing a sound detected control signal in response to an analog sound signal of a certain amount of energy;

(c) a first analog channel having a sound detected input coupled to the sound detected output, an analog sound signal input coupled to the amplified analog sound signals, and a first extracted feature output, the first analog channel normally being in a power off state and powering up in response to the sound detected control signal, the first analog channel producing a first frame based sparse feature signal on the first extracted feature output based on a first sparse feature in response to the amplified analog signal input;

(d) a second analog channel, connected in parallel with the first analog channel, having a sound detected input coupled to the sound detected output, an analog sound signal input coupled to the amplified analog sound signals, and a second extracted feature output, the second analog channel normally being in a power off state and powering up in response to the sound detected control signal, the second analog channel producing a second frame based sparse feature signal on the second extracted feature output based on a second sparse feature, which is different from the first sparse feature, in response to the amplified analog signal input;

(e) event trigger logic having inputs coupled to the first and second extracted feature outputs and an event trigger output, the event trigger logic normally being in a power off state and powering up in response to the sound detected control signal, the event trigger logic producing an event trigger signal in response to at least one of the frame based sparse feature signals indicating a starting point; and

(f) classifier circuitry having inputs coupled to the first and second extracted feature outputs and the event trigger output, the classifier circuitry normally being in a powered off state and powering up in response to the event trigger signal, the classifier circuitry performing sound recognition based on the analog sound signals.

6. The sensor of claim 5 in which the first sparse feature is one of zero crossing information, total energy values, band pass energy values, high pass energy values, or low pass energy values and the second sparse feature is another one of zero crossing information, total energy values, band pass energy values, high pass energy values, or low pass energy values.

7. The sensor of claim 5 in which the first and second analog channels produce the frame based sparse feature signals from which speech content cannot be reconstructed.

8. The sensor of claim 5 in which the classifier circuitry performs sound recognition of a portion of a spoken command of one second in duration.