Speech enhancement and source separation supported by negative beamforming filtering
Agustín Álvarez-Marquina, Pedro Gómez‐Vilda, Victor M. Garca Nieto, R. Martínez, V. Rodellar · 2003
Beamforming filtering techniques are well known for their angular selectivity in speech enhancement applications with noisy backgrounds. Besides, first order differential null beamformers achieve an important advantage compared with classical array structures: the lower number of signal sensors (microphones) required (see Elko, G.W., 1996). The scheme that we propose combines two techniques: FOD null beamforming and spectral subtraction. These methods may be used to eliminate or enhance one or several signals/sources coming from different locations through the use of a simple FOD array. The fundamentals of negative (null) array processing are reviewed and some results obtained in a real environment are also included. The simplicity of the audio acquisition equipment allows building end-user products at a reasonable cost. Applications of this system may be found in speech enhancement for communication channels and robust speech recognition methods for surveillance systems or demotic control.