Research Of Frequency Invariant Beamformers

Array signal processing is used widely in radar, sonar, communication, seismicprospecting, astronomy and medical diagnosis, etc. Along with the development of the speechprocessing and array signal processing, the microphone array signal processing technologyfuse the space-time information of speech signals to control the beamforming which have thefunction of improving the useful signal gain and suppressing the interference signal.Beamforming is an important part of array signal processing, so several classic methodsof beamforming are introduced in this paper. This paper expounds the existing frequencyinvariant beamforming methods, and the frequency invariant beamforming theories and methodsare summed up and analyzed. These beamforming methods include the minimax value method, thesubband decomposition method, bessel function method, etc. The paper focus on the Fouriertransform FIB. These frequency invariant beamforming methods are compared incomputational complexity, realizability and economic benefit.The simulation experiments onthe classical Fourier transform FIB further demonstrate the effective and the disadvantages inthe low frequency.A non-uniform fourier transformation method for frequency invariant beamformingbased on microphone array speech processing is proposed, which will solve the problem thatFIB cannot achieve a good frequency invariant property for the lower frequencies. Thismethod adopted non-uniform sampling to increase points in the low frequencies, structuringvandermonde matrix which accorded with non-uniform Fourier inverse transformation. Thenthe results were cut out using optimal symmetric window function. Finally, a space-time filterwas got. The simulation experiments show that not only the beam diagram obtainedapproximates to expected beam response, but also the number of sensors is reduced. It canthus control the cost effectively to be used in the practical application.