A Speech Enhancement in Diffuse Noise Field using Minimum Variance Distortionless Response Filter

Array signal processing is an important branch in the field of signal processing. It has been widely used in many military and national economic fields such as radar, communications, sonar, navigation, voice signal processing, seismic monitoring, and biomedical engineering. Adaptive beamforming is an important research content in array signal processing. It is essential to perform spatial filtering through adaptive weighting of each array element to achieve the purpose of enhancing the desired signal, suppressing interference and reducing noise signals. Minimum variance distortion response (MVDR) is a commonly used algorithm. It restricts the array gain to 1 in the desired signal direction and minimizes the output power of the array to achieve the purpose of suppressing interference. However, in real-life environment, most processing speech system are affected by various types of noise, like coherence, incoherence, diffuse and ambient noise. These unwanted factors strong degrade performance of designed system. The quality and intelligibility, which one of the most important measurement of output signal, decreases when surrounding environment is too complex. In this paper, the author proposed an improved technology to enhance algorithm MVDR in diffuse noise field. The results illustrated that, the suggested algorithm allowed reducing speech distortion from 5.1÷15.5 (dB). The promising method can be further combined with many speech applications.