O. Thiergart, M. Taseska, and E.A.P. Habets
Published in the IEEE Transactions on Audio, Speech, and Language Processing, Vol. 22, Issue 12, pp. 2182-2196, Dec. 2014.
Extracting desired source signals in noisy and reverberant environments is required in many hands-free communication systems. In practical situations, where the position and number of active sources may be unknown and time-varying, classical filters do not provide sufficiently good performance. Recently, informed spatial filters have been introduced that incorporate almost instantaneous parametric information on the sound field, thereby enabling adaptation to new acoustic conditions and moving sources. In this contribution, we propose a spatial filter which generalizes the recently proposed informed linearly constrained minimum variance (LCMV) filter and informed minimum mean square error (MMSE) filter. The filter uses multiple direction-of-arrival (DOA) estimates and second-order statistics of the noise and diffuse sound. To estimate those statistics, an optimal diffuse power estimator is proposed that outperforms existing methods. Extensive performance evaluation demonstrates the effectiveness of the proposed filter in dynamic acoustic conditions.
Details of the simulation setup are described in . A shoebox room (6 x 5 x 3.5m) with RT60 = 275ms was simulated. The sound was captured with a non-uniform linear array with M=6 omnidirectional microphones. Two speech sources were located in front of the array a distance of 1.7m. Source A (male speaker) represents the desired source at an angle of 109 degrees. Source B (female speaker) represents an interferer who is active from different positions. Moth speakers are active at the same time (double talk). Microphone self-noise was added to the microphone signals (34dB segSNR).
We use recently proposed informed spatial filters to extract the desired speaker and attenuate the interferer. The informed LCMV filter  provides a distortionless response for the direct sound while minimizing the noise and reverberation. The informed MMSE filter  minimizes the means-squared error of the output signal. The informed PMMW filter represents a generalization of the informed LCMV and informed MMSE filter.
For comparison, we consider two spatial filters with fixed look direction towards the desired speaker, one filter maximizing the white noise gain (WNG) and one filter maximizing the directivity index (DI).
Example 1: Interferer moving slowly:
Example 2: Interferer jumping to different positions:
Details of the measurement setup are described in . The measurement setup was almost identical to the simulation setup described above, but with a higher reverberation time (RT60 = 390ms). The same spatial filters as described above were computed. We provide all signals used in the MUSRA listening test in .
Example 3: Interferer jumping to different loudspeaker positions
O. Thiergart, M. Taseska and E.A.P. Habets, "An informed parametric spatial filter based on instantaneous direction-of-arrival estimates," IEEE/ACM Transactions on Audio, Speech and Language Processing, Vol. 22, Issue 12, pp. 2182-2196, Dec. 2014.
O. Thiergart and E.A.P. Habets, "An Informed LCMV Filter based on Multiple Instantaneous Direction-of-Arrival Estimates", in IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2013.