Audio signal detector
An audio signal detector comprising: an input adapted to receive a transducer signal; an output to provide an output signal indicative of the presence of a voice component in the transducer signal; and a processor circuit adapted to estimate a noise characteristic of the transducer signal; detect a voice component of the transducer signal; adjust one or more adjustable voice-activity parameters for detecting the voice component, wherein the said one or more adjustable voice-activity parameters are adjusted based on the noise characteristic; and generate the output signal.
1. An audio signal detector comprising:
an input to receive a transducer signal from a transducer;
an output to provide an output signal indicative of the presence of a voice component in the transducer signal; and
a processor circuit adapted to:
estimate a noise characteristic of the transducer signal;
detect a voice component of the transducer signal;
adjust one or more adjustable voice-activity parameters for detecting the voice component, wherein the one or more adjustable voice-activity parameters are adjusted based on the noise characteristic of the transducer signal; and
generate the output signal,
wherein the noise characteristic of the transducer signal comprises at least one of a noise intensity level and a noise type.
2. The audio signal detector as claimed in claim 1 , wherein the processor circuit comprises: a first estimator to estimate the noise characteristic of the transducer signal, and a second estimator to detect the voice component and to generate the output signal.
3. The audio signal detector as claimed in claim 2 , wherein the second estimator is configured to detect a start and an end of the voice component; and wherein the output signal is a binary signal configured to
transition from a first state to a second state when the start of the voice component is detected; and
transition from the second state to the first state when the end of the voice component is detected.
4. The audio signal detector as claimed in claim 3 , wherein the audio signal detector further comprises an adjuster configured to adjust the one or more voice-activity parameters based on the noise characteristic.
5. The audio signal detector as claimed in claim 4 , wherein the adjuster is configured to
adjust the one or more voice-activity parameters when the output signal is in the first state.
6. The audio signal detector as claimed in claim 2 , wherein
the second estimator comprises an integrator configured to compute a first energy integral of the transducer signal over a first time interval; and
the second estimator is configured to detect the presence of the voice component in the transducer signal based on the first energy integral and on the one or more adjustable voice-activity parameters.
7. The audio signal detector as claimed in claim 6 , wherein
the one or more adjustable voice-activity parameters comprise one or more start parameters and one or more end parameters; and
the second estimator is configured to detect the start of the voice component and the end of the voice component based on the one or more start parameters and on the one or more end parameters, respectively.
8. The audio signal detector as claimed in claim 7 , wherein
the one or more start parameters comprise a start threshold;
the one or more end parameters comprise an end threshold; and
the second estimator is configured to:
detect the start of the voice component upon identifying that the output signal is in the first state and the first energy integral rises above the start threshold; and
detect the end of the voice component upon identifying that the output signal is in the second state and the first energy integral falls below the end threshold.
9. The audio signal detector as claimed in claim 1 , wherein
the one or more voice-activity parameters comprise one or more duration parameters; and
the second estimator is configured to determine that the voice component has paused but not ended according to at least of the one or more duration parameter.
10. The audio signal detector as claimed in claim 6 , wherein
the first estimator comprises an integrator configured to compute a second energy integral of the transducer signal over a second time interval, the second time interval being longer than the first time interval; and
the first estimator being configured to determine the noise characteristic of the transducer signal based on the second energy integral.
11. The audio signal detector as claimed in claim 4 , wherein
the adjuster comprises: a look-up table comprising one or more noise characteristic values of the noise characteristic and, for each noise characteristic value, a corresponding set of voice-activity parameter values; and
the adjuster being configured to select the set of the voice-activity parameter values associated with the current noise characteristic value.
12. The audio signal detector as claimed in claim 1 , wherein the noise type comprises at least one of music in the background, physical activity, nearby voices and/or city traffic; and wherein the processor circuit is adapted to:
estimate the noise intensity level over a first time interval; and
determine whether the noise intensity level falls into one of a plurality of predetermined noise intensity ranges.
13. The audio signal detector as claimed in claim 1 , wherein the transducer is a bone conduction sensor.
14. The audio signal detector as claimed in claim 1 , wherein the transducer has a bandwidth ranging from about 600 Hz to about 2 KHz.
15. The audio signal detector as claimed in claim 1 , wherein the audio signal detector further comprises a pre-processor coupled to the processor circuit, the pre-processor being configured to pre-process the transducer signal.
16. An automatic speech recognition (ASR) system, the system comprising:
a microprocessor adapted to execute an ASR software;
an audio signal detector; and
a controller adapted to receive an output signal from the audio signal detector and to generate a control signal to control the microprocessor;
wherein
the audio signal detector comprises:
an input adapted to receive a transducer signal;
an output to provide the output signal, the output signal being indicative of the presence of a voice component in the transducer signal; and
a processor circuit adapted to:
estimate a noise characteristic of the transducer signal;
detect a voice component of the transducer signal,
adjust one or more adjustable voice-activity parameters for detecting the voice component, wherein the one or more adjustable voice-activity parameters are adjusted based on the noise characteristic of the transducer signal; and
generate the output signal,
wherein the noise characteristic of the transducer signal comprises at least one of a noise intensity level and a noise type.
17. The ASR system as claimed in claim 16 , wherein the processor circuit comprises a first estimator to estimate the noise characteristic of the transducer signal, and a second estimator to detect the voice component and to generate the output signal; the second estimator being configured to detect a start and an end of the voice component; and the controller being configured to:
activate the microprocessor when the second estimator detects the start of the voice component; and
deactivate the microprocessor when second estimator detects the end of the voice component.
18. A user device comprising the ASR system as claimed in claim 16 .
19. A method of detecting an audio signal, the method comprising:
receiving a transducer signal;
estimating a noise characteristic of the transducer signal;
detecting a voice component of the transducer signal;
providing one or more adjustable voice-activity parameters for detecting the voice component;
adjusting the one or more voice-activity parameters based on the noise characteristic of the transducer signal; and
generating an output signal indicative of the presence of the voice component of the transducer signal,
wherein the noise characteristic of the transducer signal comprises at least one of a noise intensity level and a noise type.
20. A non-transitory computer readable medium comprising instructions that, when executed by a computer, enable it to detect an audio signal by:
receiving a transducer signal;
estimating a noise characteristic of the transducer signal;
detecting a voice component of the transducer signal,
providing one or more adjustable voice-activity parameters for detecting the voice component;
adjusting the one or more voice-activity parameters based on the noise characteristic of the transducer signal; and
generating an output signal indicative of the presence of the voice component in the transducer signal,
wherein the noise characteristic of the transducer signal comprises at least one of a noise intensity level and a noise type.