Sound Source Location Prediction Method Based on Broad Learning

In sound source localization technology, the accuracy of localization algorithms has always been a core issue in this field. This paper proposes a microphone array sound source location algorithm based on broad learning algorithm for the Time Difference of Arrival ( TDOA) technology in sound source localization (Dynamic), which solves the problems of low positioning accuracy, large computational complexity and low noise resistance in traditional localization algorithms. Unlike the structural form of deep learning, the broad learning system (BLS) is more inclined to construct the network in the "width" direction. Firstly, the audio signals collected by the microphone array are reprocessed, and the Generalized Cross-Correlation (GCC) function is mapped to the feature nodes and enhancement nodes of the broad learning system respectively. Secondly, the feature nodes and enhancement nodes are used as inputs to the neural network to construct the framework of the model. Finally, predicting the position of the sound source. We compare our method with Oneshot algorithm (broad learning without increment) and BP neural network algorithm. The experimental results show that this algorithm not only does not significantly increase the complexity of the algorithm, but also has good positioning performance.