Gökhan Ince, Kazuhiro Nakadai, Tobias Rodemann, Hiroshi Tsujino, Jun-ichi Imura, "Ego-motion Noise Cancellation of a Robot using Missing Feature Masks", Applied Intelligence, vol. 34, no. 3, pp. 360-371, 2011.

This paper describes an architecture that enhances a robot with the capability of performing au- tomatic speech recognition by cancelling the ego noise, even while the robot is moving. The system consists of three blocks: (1) a multi-channel noise reduction block comprising consequent stages of microphone-array-based sound localization, geometric source separation and post- ltering, (2) a single-channel noise reduction block uti- lizing template subtraction and (3) a speech recogni- tion block. In this work, we specically investigate a missing feature theory based automatic speech recogni- tion (MFT-ASR) approach in block (3). It makes use of spectro-temporal elements that are derived from (1) and (2) to measure the reliability of the audio features and generates masks to lter unreliable speech features. We evaluate the proposed system on a robot using word error rates. Furthermore, we present a detailed analysis of recognition accuracy to determine optimal parame- ters. Proposed MFT-ASR implementation attains sig- nicantly higher recognition performance compared to the performance of single or multi-channel noise reduc- tion methods.