Methods and apparatus to operate an audience metering device with voice commands
View Patent ↗Methods and apparatus to operate an audience metering device with voice commands are described herein. In an example method, at least one of a television program audio signal or a voice command from an audience member is transduced into an audio input signal. Based on the audio input signal and a television audio line signal, a residual audio signal is generated. One or more vectors from the residual audio signal are extracted. Based on the one or more vectors extracted from the residual audio signal, the voice command is identified.
1. A method to identify audience members based on voice, the method comprising:
obtaining an audio input signal including a television program audio signal and a human voice signal;
receiving a television audio line signal from an audio output line of a monitored television;
processing the television audio line signal with a filter having adaptive weights to generate a delayed and attenuated television line signal;
subtracting the delayed and attenuated television line signal from the audio input signal to develop a residual audio signal;
identifying a person that spoke to create the human voice signal based on the residual audio signal; and
logging an identity of the person as an audience member.
2. A method as defined in claim 1 wherein identifying the person further comprises:
extracting one or more identifiers from the residual audio signal; and
comparing the one or more identifiers extracted from the residual audio signal to a database of stored identifiers.
3. A method as defined in claim 2 wherein extracting one or more identifiers from the residual audio signal comprises:
shifting a plurality of discrete samples of the residual audio signal into a buffer to form a digital audio line sample;
performing a fast fourier transform on the digital audio line sample in the buffer to create an FFT spectrum;
defining frequency bands in the FFT spectrum;
transforming the frequency bands into coefficients; and
selecting a set of the coefficients corresponding to vocal tract resonances as the one or more identifiers.
4. A method as defined in claim 3 further comprising padding the buffer to increase a length of the digital audio line sample.
5. A method as defined in claim 3 further comprising performing a windowing function to reduce block boundary effects.
6. A method as defined in claim 3 wherein the digital audio line sample includes the plurality of discrete samples of the residual audio signal and a plurality of discrete samples of an earlier residual audio signal.
7. A method as defined in claim 3 wherein the frequency bands are overlapping mel filter frequency bands.
8. A method as defined in claim 3 wherein transforming the frequency bands into the coefficients comprises transforming log spectral energy values associated with the frequency bands by a discrete cosine transform.
9. A method as defined in claim 3 wherein the set of the coefficients further comprises a coefficient substantially corresponding to a total energy of the digital audio line sample.
10. A method as defined in claim 2 wherein extracting the one or more identifiers from the residual audio signal comprises extracting one or more mel frequency cepstral coefficients (MFCC) vectors from the residual audio signal.
11. A method as defined in claim 1 wherein processing the television audio line signal with the filter having adaptive weights to generate the delayed and attenuated television line signal further comprises:
sampling the television audio line signal into samples;
inputting the samples into a delay line;
multiplying the samples with respective ones of the adaptive weights to generate a plurality of weighted samples; and
summing the weighted samples.
12. A method as defined in claim 11 further comprising adjusting values of the adaptive weights.
13. A method as defined in claim 11 wherein adjusting the values of the adaptive weights comprises summing respective ones of the values of the adaptive weights with a respective one of a set of adjustment factors, each of the adjustment factors comprising a product of: (a) the residual audio signal at a first time, (b) a learning factor, and (c) a respective sample of the television audio line signal at the first time.
14. A method as defined in claim 1 wherein processing the television audio line signal with the filter having adaptive weights to generate the delayed and attenuated television line signal comprises minimizing a least mean squared error.
15. A machine accessible medium storing instructions, which when executed, cause a machine to:
obtain an audio input signal including a television program audio signal and a human voice signal;
receive a television audio line signal from an audio output line of a monitored television;
process the television audio line signal with a filter having adaptive weights to generate a delayed and attenuated television line signal;
subtract the delayed and attenuated television line signal from the audio input signal to develop a residual audio signal;
identify a person that spoke to create the human voice signal based on the residual audio signal; and
log an identity of the person as an audience member.
16. An audience measurement device to identify audience members based on voice, the device comprising:
a transducer to obtain an audio input signal including a television program audio signal and a human voice signal;
a filter having adaptive weights to receive a television audio line signal from an audio output line of a monitored television, and to process the television audio line signal to generate a delayed and attenuated television line signal;
a difference detector to subtract the delayed and attenuated television line signal from the audio input signal to develop a residual audio signal;
an identification device to identify a person that spoke to create the human voice signal based on the residual audio signal; and
a memory to log an identity of the person as an audience member.
17. An apparatus as defined in claim 16 wherein the identification device further comprises:
a feature extractor to extract one or more identifiers from the residual audio signal; and
a matcher to compare the one or more identifiers extracted from the residual audio signal to a database of stored identifiers.
18. An apparatus as defined in claim 17 wherein the feature extractor further comprises:
a buffer to receive a plurality of discrete samples of the residual audio signal to form a digital audio line sample; and
a processor to fast fourier transform the digital audio line sample in the buffer to create an FFT spectrum, to define frequency bands in the FFT spectrum, to transform the frequency bands into coefficients, and to select a set of the coefficients corresponding to vocal tract resonances as the one or more identifiers.
19. An apparatus as defined in claim 18 wherein the processor pads the buffer to increase a length of the digital audio line sample before transforming the digital audio line sample.
20. An apparatus as defined in claim 18 wherein the processor performs a windowing function to reduce block boundary effects associated with the fast fourier transform.
21. An apparatus as defined in claim 18 wherein the digital audio line sample includes the plurality of discrete samples of the residual audio signal and a plurality of discrete samples of an earlier residual audio signal.
22. An apparatus as defined in claim 18 wherein the frequency bands are overlapping mel filter frequency bands.
23. An apparatus as defined in claim 18 wherein the processor transforms the frequency bands into the coefficients by transforming log spectral energy values associated with the frequency bands with a discrete cosine transform.
24. An apparatus as defined in claim 18 wherein the set of the coefficients further comprises a coefficient substantially corresponding to a total energy of the digital audio line sample.
25. An apparatus as defined in claim 16 wherein the filter further comprises:
an analog-to-digital converter to sample the television audio line signal into samples;
a delay line to receive the samples;
a multiplier to multiply the samples with respective ones of the adaptive weights to generate a plurality of weighted samples; and
an adder to sum the weighted samples.
26. An apparatus as defined in claim 25 further comprising a weight adjuster to adjust values of the adaptive weights.
27. An apparatus as defined in claim 26 wherein the weight adjuster adjusts the values of the adaptive weights by summing respective ones of the values of the adaptive weights with a respective one of a set of adjustment factors, each of the adjustment factors comprising a product of: (a) the residual audio signal at a first time, (b) a learning factor, and (c) a respective sample of the television audio line signal at the first time.