Spatial Resolution Enhancement via Hybrid Estimation Approaches

Since the beamforming concept was first introduced to acoustic engineers approximately 10 years ago, the acoustic phased array has become one of the most popular techniques used to identify noise sources. Generally speaking, the spatial resolution of the phased array is proportional to the number of microphones and the array size. With more microphones or a larger array, better spatial resolution is achieved. However, to achieve better spatial resolution for a given microphone array structure, the so-called CLEAN method was recently introduced. By applying the traditional non-parametric estimation method, such as delay-and-sum, the CLEAN algorithm first computes the largest noise and removes it from the raw data samples. Then, the program estimates a new peak noise, based on the “cleaned” data samples. The process is repeated until the remaining data samples become random noise. The CLEAN method works well if the noise sources are far apart. However, if the noise sources are close to each other, the non-parametric peak estimation fails to distinguish different sources. To overcome this limitation, this paper details the application of the MUSIC and ESPRIT parametric algorithms for initial peak detection before performing the CLEAN method. As proved by our real-world experiments, the proposed hybrid approach can significantly improve the spatial resolution obtained by the original CLEAN method.