Beam formation based on GSC microphone Array (brief)

This paper refers to some information and discussion on the Internet, and expresses our thanks.

Microphone array beamforming has been widely researched for wireless communication, speech recognition, speech enhancement, hearing aids and so on. Compared with other domain array processing, the microphone array processing has its own characteristics determined by the difficulties. This is due to the statistical non-stationary nature of speech signals, and often with complex echo backgrounds. In various beamforming methods, adaptive beamforming is more advantageous than conventional fixed beamforming, and adaptive beamforming has good performance in suppressing interference and noise. The Griffiths-jim beamforming (GJBF) is a widely researched and applied adaptive beamforming, but the GJBF is very sensitive to the guidance vector, and the target signal cancellation occurs under the condition of the vector error. In the actual situation, the guide vector error is unavoidable, and the main causes of the guide vector error are:

1. The microphone array itself cannot be ideal, 2. The gain between microphones is different, 3. The influence of the source direction (DOA). For example, for wireless communications such as car phones, the directional vector error caused by the instability of DOA is the main reason for the target signal cancellation.

Since the guidance vector error is unavoidable, a variety of robust beamforming is proposed, and their focus is on increasing tolerance to the vector error of guidance. However, these beamforming are often at the expense of signal noise interference ratios, and the number of microphone arrays required increases. The beam-forming device quoted in this paper is a robust beamforming based on generalized sidelobe Canceller (Hoshuyama), and its blocking filter and multi-input canceller respectively apply the coefficient constrained Adaptive filter (CCAF) and the Leak Detection Adaptive filter (LAF). The main features of this beam-forming device are:

1. Can provide sufficient tolerance for large target signal direction error, the maximum target signal direction error azimuth can be defined by the user; 2. Suitable for small microphone array; 3. Good anti-jamming noise performance.

A sidelobe Canceller for receiving M-mic signals is shown in the figure. Prior to setting up an equal-interval wideband linear microphone array, all kinds of signals, including the target signal and noise, are transmitted in a plane wave manner, and the exact direction of the target signal propagation is assumed to be known.

a wideband sidelobe Canceller includes a beamforming (FBF), a blocking filter (BM), and a multi-input canceller (MC). The fixed beamforming enhances the target signal and D (k) is the output of the fixed beamforming at the sampling time K, and Xm (k) is the output signal of the M array. The multi-input canceller can adaptively fix the output delay signal of the beam-forming device minus the component associated with the YM (k) output signal of the blocking filter, where q is the delay amount. The blocking filter is a kind of spatial impedance band filter, which can suppress the target signal and pass the interfering signal. If the input signal ym (k) of the multiple input Canceller contains only the interfering signal, the multi-input canceller can suppress the interference signal and extract the target signal. However, if the YM (K) contains a target signal component, the target signal is also eliminated in the multi-input canceller.

In some simple wideband sidelobe Canceller, the blocking filter is very sensitive to the guide vector error and it is very easy to leak the target. In the actual work, it is impossible to know exactly the direction of the target signal, so the guide vector error is unavoidable. Therefore, the elimination of signals is a very important issue.

There are many signal processing methods to avoid the elimination of target signals. Some robust beamforming algorithms introduce constraints to the adaptive algorithm for multiple input canceller. Adaptive algorithms with leakage, noise, and norm constraints can limit unwanted signal cancellation. These robust beamforming can be a target signal pass when there is a small pointing vector error. However, when there is a large target signal error, the elimination of the interference signal will also be limited.

Some robust beamforming uses an improved spatial filter in blocking filtering. This filter eliminates the presence of a pointing vector error. However, this filter can only be used when there is a small target direction error. When there is a large target direction error, the spatial filter loses the degree of freedom to suppress interference. This loss of freedom can greatly reduce the performance of suppressing interference.

Target tracking or calibration is another way to form a robust beamforming. This allows large directional vector errors to exist without losing degrees of freedom without compromising the performance of interference suppression. However, if the target signal is a pulse signal, error tracking occurs. Furthermore, the use of matrix theory for accurate target tracking requires a large amount of computation.