Time-based frequency tuning of analog-to-information feature extraction

• US 10,373,608 B2
• Filed: 10/22/2015
• Issued: 08/06/2019
• Est. Priority Date: 10/22/2015
• Status: Active Grant

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Zhang, Zhenyong, Ma, Wei
• Primary Examiner(s): Yen, Eric
• Application Number:
  US14/920,210
• Publication Number:
  US 20170116980A1
• Time in Patent Office: 1,384 Days
• 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 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
• CPC Class Codes: G10L 21/0224 Processing in the time domain
• CPC Class Codes: G10L 21/0264 characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques
• CPC Class Codes: G10L 25/09 the extracted parameters being zero crossing rates
• CPC Class Codes: G10L 25/21 the extracted parameters being power information

Abstract

A sound recognition system including time-dependent analog filtered feature extraction and sequencing. An analog front end (AFE) in the system receives input analog signals, such as signals representing an audio input to a microphone. Features in the input signal are extracted, by measuring such attributes as zero crossing events and total energy in filtered versions of the signal with different frequency characteristics at different times during the audio event. In one embodiment, a tunable analog filter is controlled to change its frequency characteristics at different times during the event. In another embodiment, multiple analog filters with different filter characteristics filter the input signal in parallel, and signal features are extracted from each filtered signal; a multiplexer selects the desired features at different times during the event.

Claims

1. An audio recognition circuit, comprising:

a first analog filter for filtering an analog signal using a first frequency characteristic over a first interval of a particular duration;

a second analog filter for filtering the analog signal using a second frequency characteristic different from the first frequency characteristic over a second interval of the particular duration differing from the first interval of the particular duration;

first feature extraction circuitry for extracting a first analog signal feature from the analog signal filtered by the first analog filter and corresponding to the first interval;

second feature extraction circuitry for extracting a second analog signal feature from the analog signal filtered by the second analog filter and corresponding to the second interval;

a multiplexer for outputting a feature sequence comprising the extracted first and second analog signal features;

an analog-to-digital converter for digitizing the feature sequence; and

an event trigger for issuing an event trigger signal responsive to the feature sequence matching a particular feature sequence according to a matching criterion.

2. The circuit of claim 1, further comprising:

digital sound recognition circuitry for performing digital sound recognition on the digitized feature sequence responsive to the event trigger signal.

3. The circuit of claim 1, wherein the first feature extraction circuitry comprises a first zero crossing counter for detecting a number of times the analog signal crosses a first threshold level over the first interval, and the second feature extraction circuitry comprises a second zero crossing counter for detecting a number of times the analog signal crosses a second threshold level over the second interval.

4. The circuit of claim 1, wherein the first feature extraction circuitry comprises a first integrator for measuring a first total energy of the analog signal over the first interval, and the second feature extraction circuitry comprises a second integrator for measuring a second total energy of the analog signal over the second interval.

5. The circuit of claim 1, wherein the first and second analog filters are tunable in response to one or more control signals.

6. The circuit of claim 1, wherein the first frequency characteristic comprises a low pass filter characteristic with a first cutoff frequency, and the second frequency characteristic comprises a low pass filter characteristic with a second cutoff frequency different from the first cutoff frequency.

7. The circuit of claim 1, wherein the event trigger comprises:

circuitry for comparing the feature sequence with the particular feature sequence according to the matching criterion.

8. The circuit of claim 1, wherein the event trigger comprises:

a communications link for communicating the feature sequence to a database server; and

circuitry for issuing the event trigger responsive to receiving a signal, from the database server over the communications link, indicating that the matching criterion is met by the feature sequence.