Method for acquiring noise-refined voice signal, and electronic device for performing same
According to various embodiments, an electronic device includes a plurality of microphones, and a processor electrically connected to the plurality of microphones, wherein the processor may obtain audio signals through the plurality of microphones, estimate a probability of existence of a voice signal included in the obtained audio signals, obtain correlation information between the audio signals based on the probability of existence of the voice signal and/or the obtained audio signals, obtain voice blocking information based on the correlation information or a direction of arrival (DOA) estimation, obtain a first signal among the audio signals based on the audio signals, the correlation information, and the voice blocking information, obtain a second signal including the voice signal among the audio signals, and obtain a noise-removed voice signal by removing the first signal from the second signal. In addition, it is possible to implement various embodiments understood through the disclosure.
1. An electronic device comprising:
a plurality of microphones; and
a processor electrically connected to the plurality of microphones,
wherein the processor is configured to:
obtain audio signals through the plurality of microphones,
estimate a probability of existence of a voice signal included in the obtained audio signals,
obtain estimated voice signals based on the audio signals and the probability of existence of the voice signal,
obtain correlation information between the audio signals based on the probability of existence of the voice signal and/or the obtained audio signals, wherein the correlation information includes a covariance matrix between the estimated voice signals that correspond to the plurality of microphones, respectively,
obtain voice blocking information based on the correlation information or a direction of arrival (DOA) estimation,
obtain a first signal among the audio signals based on the audio signals, the correlation information, and the voice blocking information,
obtain a second signal including the voice signal among the audio signals, and
obtain a noise-removed voice signal by removing the first signal from the second signal.
2. The electronic device of claim 1 , wherein the processor is configured to:
obtain a blocking matrix based on the correlation information and the voice blocking information, and
obtain the first signal based on the audio signals and the blocking matrix.
3. The electronic device of claim 1 , wherein the correlation information includes a covariance matrix between the audio signals that correspond to the plurality of microphones, respectively.
4. The electronic device of claim 1 , wherein the processor is configured to obtain the voice blocking information based on the covariance matrix between the estimated voice signals.
5. The electronic device of claim 4 , wherein the processor is configured to obtain the voice blocking information based on a column vector of the covariance matrix between the estimated voice signals.
6. The electronic device of claim 1 , wherein the processor is configured to:
obtain estimated noise signals based on the audio signals and the probability of existence of the voice signal, and
obtain the voice blocking information based on a covariance matrix between the estimated noise signals.
7. The electronic device of claim 6 , wherein the processor is configured to obtain the voice blocking information based on an eigen vector of the covariance matrix between the estimated noise signals.
8. The electronic device of claim 6 , wherein the processor is configured to obtain the voice blocking information based on a column vector of the covariance matrix between the estimated noise signals.
9. The electronic device of claim 1 , wherein the processor is configured to perform the direction of arrival estimation based on at least a difference between times taken for the voice signal to reach the plurality of microphones.
10. A method of obtaining a noise-removed voice signal among audio signals by an electronic device, the method comprising:
obtaining the audio signals;
estimating a probability of existence of a voice signal;
obtaining estimated voice signals based on the audio signals and the probability of existence of the voice signal;
obtaining correlation information based on the probability of existence of the voice signal and/or the obtained audio signals, wherein the correlation information includes a covariance matrix between the estimated voice signals that correspond to a plurality of microphones, respectively;
obtaining voice blocking information based on the correlation information or a direction of arrival (DOA) estimation;
obtaining a first signal among the audio signals based on the audio signals, the correlation information, and the voice blocking information;
obtaining a second signal including the voice signal among the audio signals; and
obtaining the noise-removed voice signal by removing the first signal from the second signal.
11. The method of claim 10 , wherein the obtaining of the first signal includes:
obtaining a blocking matrix based on the correlation information and the voice blocking information; and
obtaining the first signal based on the audio signals and the blocking matrix.
12. The method of claim 10 , wherein the obtaining of the correlation information includes:
obtaining a covariance matrix between the audio signals.
13. The method of claim 10 , wherein the obtaining of the voice blocking information includes:
obtaining the voice blocking information based on the covariance matrix between the estimated voice signals.