The method of claim 5, wherein correlation is determined by calculating a short time correlation value or a short time coherence value of the up-mixed audio signal.ħ. The method of claim 1, further comprising: determining correlations between the up-mixed audio signal and either of the left stereo audio signal and the right stereo audio signal and controlling de-correlation based on the determined correlations between the up-mixed audio signal and either of the first and second audio input signals.Ħ. The method of claim 1, wherein de-correlating of the up-mixed audio signal includes transforming the up-mixed audio signal from a time domain to a frequency domain to generate a multiplicity of sub-band signals, multiplying each of the sub-band signals with a corresponding controlled modulation signal to generate modulated sub-band signals, and transforming the modulated sub-band signals from the frequency domain into the time domain to provide the de-correlated up-mixed audio signal.ĥ. The method of claim 2, wherein the at least one time-varying parameter is predetermined or controllable.Ĥ. The method of claim 1, wherein de-correlating of the up-mixed audio signal includes generating the de-correlated up-mixed audio signal by applying a time-varying all-pass filtering to the up-mixed audio signal, the time-varying all-pass filtering being based on at least one time-varying parameter.ģ. A method for multi-channel audio signal processing, comprising: receiving a left stereo audio signal from a first channel, and receiving a right stereo audio signal from a second channel up-mixing the left stereo audio signal and the right stereo audio signal to generate an up-mixed audio signal for a third channel de-correlating the up-mixed audio signal from the left stereo audio signal and the right stereo audio signal to generate a de-correlated up-mixed audio signal providing the left stereo audio signal to a first loudspeaker to generate a first sound signal, providing the right stereo audio signal to a second loudspeaker to generate a second sound signal, and providing the de-correlated up-mixed audio signal to a third loudspeaker to generate a third sound signal picking up the first, second and third sound signals with a microphone to generate a microphone output signal and adaptively filtering the microphone output signal with an acoustic echo canceller based on the left stereo audio signal, the right stereo audio signal and the de-correlated up-mixed audio signal to generate an echo compensated microphone signal.Ģ. Experiments show that the computational cost grows in proportion to the linear order of the reverberation time and that our method improves the word correctness of automatic speech recognition by 10 to 20 points in a RT₂₀= 670 ms reverberant environment.1. In this letter, we reduce the computational complexity to the linear order of the reverberation time by using two techniques: (1) a separation model based on the independence of delayed observed signals with MINT and (2) spatial sphering for preprocessing. Therefore, the main issue in dereverberation is to reduce the high computational cost of ICA. A naive implementation of this method is computationally expensive, because its time complexity is the second order of reverberation time. To extract a clean signal from the reverberant observation, we model the separation process in the short-time Fourier transform domain and apply the multiple input/output inverse-filtering theorem (MINT) to the FD-ICA separation model.
#Signal processing and acoustic echo cancellation how to#
In applying conventional FD-ICA as a preprocessing of automatic speech recognition in noisy environments, one of the most critical problems is how to cope with reverberations. We focus on frequency domain ICA (FD-ICA) because its computational cost and speed of learning convergence are sufficiently reasonable for practical applications such as hands-free speech recognition. This letter presents a new algorithm for blind dereverberation and echo cancellation based on independent component analysis (ICA) for actual acoustic signals.
0 kommentar(er)
