Spatial-feature-based acoustic scene analysis using distributed microphone array

In this paper we propose a robust and efficient method to utilize the spatial information provided by a distributed microphone array for acoustic scene analysis. In our approach, similarly to the cepstrum, which is widely used as a spectral feature, the logarithm of the amplitude in multichannel observation is converted to a feature vector by a linear orthogonal transformation. Then, the spatial information of the acoustic scene is represented in the spatial feature space. This approach does not require the positions of the microphones and is not sensitive to the synchronization mismatch of channels, both of which make the method suitable for use with a distributed microphone array. Experimental results using reallife environmental sounds show the validity of our approach even when a smaller feature dimension than the original one is used.