A neuromorphic microphone for sound localization

We propose a localization model which uses monaural spectral cues for localizing a sound source in a 1-D plane. A neuromorphic microphone is constructed to implement this model. We use the term ''neuromorphic'' because the microphone's operating principles take advantage of biologically-based monaural cues. By concentrating on monaural cues, we hope to better understand human sound localization and also build low-cost stereo capabilities into a single microphone. The Head Belated Transfer Function (HRTF) plays a critical role for human monaural localization since the shape of the external ear (pinna) spectrally shapes the sound differently for each sound source direction. Using HRTFs, humans can perceive the difference between front and back and the sound source's different elevation positions using only a single ear. The neuromorphic microphone relies on a specially shaped reflecting structure that allows echo-time processing to localize the sound. Since our recorded signal is composed of the direct sound and its echo, the sound is a simplified version of actual HRTF recordings which are composed of the direct sound and its reflections from the external ear, head, shoulder, and torso. The recorded signal from our special microphone is first processed using a gamma filter. The gamma Alter generalizes the standard transversal Alter by adding the ability to choose an optimal time-scale. Our studies have shown that the gamma filter solutions require on the order of Ave parameters while the more typical FIR filter solutions require hundreds of parameters. A multilayer perceptron neural network is then used to learn the elevation angle of the sound-allowing the microphone to correctly localize sounds. With the successful building of the special microphone, a full hardware version of a neuromorphic microphone is possible.