System and method for single channel distant speech processing
A method, computer program product, and computing system for receiving a signal from a single microphone. A plurality of modified signals may be generated from the single microphone signal, where the plurality of modified signals include at least one of: a speaker-specific signal, an acoustic parameter-specific signal, and a speech enhanced signal. Speech processing may be performed on the plurality of modified signals.
1. A computer-implemented method, executed on a computing device, comprising:
receiving a signal from a single microphone;
generating a plurality of modified signals from the signal received from the single microphone, wherein the plurality of modified signals include at least a first acoustic parameter-specific signal and a second acoustic parameter-specific signal, the first acoustic parameter-specific signal having a higher C50 value than the second acoustic parameter-specific signal;
weighting each of the plurality of modified signals, wherein the first acoustic parameter-specific signal is assigned a higher weight than the second acoustic parameter-specific signal based on the first acoustic parameter-specific signal having a higher C50 value than the second acoustic parameter-specific signal;
generating a combined modified signal by combining the plurality of modified signals based upon, at least in part, the weighting of each of the plurality of modified signals; and
performing speech recognition on the combined modified signal.
2. The computer-implemented method of claim 1 , wherein the plurality of modified signals includes a plurality of speaker-specific signals associated with a plurality of speakers.
3. The computer-implemented method of claim 2 , wherein generating the plurality of speaker-specific signals associated with the plurality of speakers includes constructing a mask for each speaker in the plurality of speakers to attenuate speech associated with other speakers in the plurality of speakers, thus defining a plurality of speaker masks.
4. The computer-implemented method of claim 3 , wherein constructing the mask for each speaker in the plurality of speakers includes constructing a quantized mask with an attenuation factor based upon, at least in part, a speaker identification confidence value.
5. The computer-implemented method of claim 1 , wherein generating the plurality of modified signals includes performing a plurality of speech enhancements on the signal received from the single microphone.
6. The computer-implemented method of claim 1 , wherein combining the plurality of modified signals includes:
extracting feature information from each modified signal in the plurality of modified signals; and
combining the extracted feature information.
7. The computer-implemented method of claim 1 , wherein performing speech recognition on the combined modified signal includes:
performing speech recognition on the combined modified signal using a multi-channel speech processing system.
8. A computing system comprising:
a memory storing programming instructions; and
a processor configured to execute the programming instructions stored by the memory, wherein the programming instructions, upon execution by the processor, cause the computing system to:
receive a signal from a single microphone;
generate a plurality of modified signals from the signal received from the single microphone, wherein the plurality of modifies signals include at least a first acoustic parameter-specific signal and a second acoustic parameter-specific signal, the first acoustic parameter-specific signal having a higher C50 value than the second acoustic parameter-specific signal;
weight each of the plurality of modified signals, wherein the first acoustic parameter-specific signal is assigned a higher weight than the second acoustic parameter-specific signal based on the first acoustic parameter-specific signal having a higher C50 value than the second acoustic parameter-specific signal;
generate a combined modified signal by combining the plurality of modified signals based upon, at least in part, the weighting of each of the plurality of modified signals; and
perform speech recognition on the combined modified signal.
9. The computing system of claim 8 , wherein the plurality of modified signals includes a plurality of speaker-specific signals associated with a plurality of speakers.
10. The computing system of claim 9 , wherein generating the plurality of speaker-specific signals associated with the plurality of speakers includes constructing a mask for each speaker in the plurality of speakers to attenuate speech associated with other speakers in the plurality of speakers, thus defining a plurality of speaker masks.
11. The computing system of claim 10 , wherein constructing the mask for each speaker in the plurality of speakers includes constructing a quantized mask with an attenuation factor based upon, at least in part, a speaker identification confidence value.
12. A computer program product comprising a non-transitory computer readable medium having a plurality of instructions stored thereon which, when executed by a processor, cause the processor to perform operations comprising:
receiving a signal from a single microphone;
generating a plurality of modified signals from the signal received from the single microphone, wherein the plurality of modifies signals include at least a first acoustic parameter-specific signal and a second acoustic parameter-specific signal, the first acoustic parameter-specific signal having a higher C50 value than the second acoustic parameter-specific signal;
weighting each of the plurality of modified signals, wherein the first acoustic parameter-specific signal is assigned a higher weight than the second acoustic parameter-specific signal based on the first acoustic parameter-specific signal having a higher C50 value than the second acoustic parameter-specific signal; and
generating a combined modifies signal by combining the plurality of modified signals based upon, at least in part, the weighting of each of the plurality of modified signals and
performing speech recognition on the combined modifies signal.
13. The computer program product of claim 12 , wherein the plurality of modified signals includes a plurality of speaker-specific signals associated with a plurality of speakers, and wherein generating the plurality of speaker-specific signals associated with the plurality of speakers includes constructing a mask for each speaker to attenuate speech associated with other speakers, thus defining a plurality of speaker masks.
14. The method of claim 1 , wherein a C50 value of the first acoustic parameter-specific signal is a ratio between sound energy received within the first fifty (50) milliseconds (ms) of a room impulse response (RIR) corresponding to the first acoustic parameter-specific signal and sound energy received after the first 50 ms of the RIR corresponding to the first acoustic parameter-specific signal.
15. The method of claim 1 , wherein the plurality of modified signals include at least two acoustic parameter-specific signals associated with different segmental signal to noise ratios (SNRs), and wherein the at least two acoustic parameter-specific signals are assigned different weights based on the different segmental SNRs associated with the at least two acoustic parameter-specific signals.
16. The method of claim 1 , wherein the plurality of modified signals include at least two acoustic parameter-specific signals having different reverberation parameters, and wherein the at least two acoustic parameter-specific signals are assigned different weights based on the different reverberation parameters associated with the at least two acoustic parameter-specific signals.
17. The method of claim 16 , wherein the reverberation parameters associated with the at least two acoustic parameter-specific signals include reverberation times of the at least two acoustic parameter-specific signals.
18. The method of claim 17 , wherein the reverberation times of the at least two acoustic parameter-specific signals correspond to a length of time required for a sound intensity of the at least two acoustic parameter-specific signals to fall below a threshold.
19. The method of claim 1 , wherein speech recognition is performed on the combined modified signal using a multi-channel speech processing system.