No abstract

No abstract

No abstract

We present a novel approach for real-time multichannel speech enhancement in environments of nonstationary noise and time-varying acoustical transfer functions (ATFs). The proposed system integrates adaptive beamforming, ATF identification, soft signal detection, and multichannel postfiltering. The noise canceller branch of the beamformer and the ATF identification are adaptively updated online, based on hypothesis test results. The noise canceller is updated only during stationary noise frames, and the ATF identification is carried out only when desired source components have been detected. The hypothesis testing is based on the nonstationarity of the signals and the transient power ratio between the beamformer primary output and its reference noise signals. Following the beamforming and the hypothesis testing, estimates for the signal presence probability and for the noise power spectral density are derived. Subsequently, an optimal spectral gain function that minimizes the mean square error of the log-spectral amplitude (LSA) is applied. Experimental results demonstrate the usefulness of the proposed system in nonstationary noise environments.

This paper presents a statistically and computationally efficient algorithm for direction finding of a single far-field source using a multi-sensor array. The algorithm extracts the azimuth and elevation angles directly from the estimated time delays between the array elements. Hence, it is referred to herein as the time delay direction finding ͑TDDF͒ algorithm. An asymptotic performance analysis, using a small error assumption, is conducted. For any 1-D and 2-D array configurations, it is shown that the TDDF algorithm achieves the Cramer Rao lower bound ͑CRLB͒ for the azimuth and elevation estimates provided that the noise is Gaussian and spatially uncorrected and that the time delay estimator achieves the CRLB as well. Moreover, with the suggested algorithm no constraints on the array geometry are required. For the general 3-D case the algorithm does not achieve the CRLB for a general array. However, it is shown that for array geometries which obey certain constraints the CRLB is achieved as well. The TDDF algorithm offers several advantages over the beamforming approach. First, it is more efficient in terms of computational load. Second, the azimuth estimator does not require the a priori knowledge of the wave propagation velocity. Third, the TDDF algorithm is suitable for applications where the arrival time is the only measured input, in contrast to the beamformer, which is not applicable in this case. INTRODUCTIONIn various applications of array signal processing such as radar, sonar, and seismology, there is a great interest in detection and localization of wideband sources. 1 The problem of estimating the direction of arrival ͑DOA͒ of wideband sources using a sensor array has been studied extensively in the literature. 2-14 A common approach 2-7 to this problem, for a single source scenario, is to use the time delay estimation between two sensors to determine the DOA. Many techniques for estimating the travel time delay between two receiving sensors have been investigated, see, e.g., Refs. 2-7. For the single source and a multi-sensor case, Hahn and Tretter 8 introduced the maximum likelihood ͑ML͒ delayvector estimator. The ML DOA estimators for the multisensor and multi-source cases have also being studied extensively. [12][13][14] It is well known 11 that the ML DOA estimator, for the single-source case with a spatially uncorrelated noise, can be realized as a focused beamformer. In this paper an alternative approach is proposed, in which the DOA is extracted directly from the estimated time delays between the array elements ͑referred to as the time delay vector͒. This approach is an extension to the multi-sensor case of the work in Refs. 10 and 11, where the DOA is extracted from the time delay between two sensors for the far-field case.The suggested time delay direction finding ͑TDDF͒ algorithm utilizes the linear relationship between the time delay vector and the DOA vector in Cartesian coordinates. This linear relationship allows a closed form estimation of the DOA vector. The transformation to polar...