Ego-Noise Reduction Using a Motor Data-Guided Multichannel Dictionary

International audienceWe address the problem of ego-noise reduction,i.e., suppressing the noise a robot causes by its own motions.Such noise degrades the recorded microphone signal massivelysuch that the robot’s auditory capabilities suffer. To suppressit, it is intuitive to use also motor data, since it providesadditional information about the robot’s joints and therebythe noise sources. We propose to fuse motor data to a recentlyproposed multichannel dictionary algorithm for ego-noise reduction.At training, a dictionary is learned that captures spatialand spectral characteristics of ego-noise. At testing, nonlinearclassifiers are used to efficiently associate the current robot’smotor state to relevant sets of entries in the learned dictionary.By this, computational load is reduced by one third in typicalscenarios while achieving at least the same noise reductionperformance. Moreover, we propose to train dictionaries ondifferent microphone array geometries and use them for egonoisereduction while the head to which the microphonesare mounted is moving. In such scenarios, the motor guidedapproach results in significantly better performance values