Virtual in-ear microphone for in-vehicle noise control based on array technology and modified zero point attraction LMS algorithms

摘要

Measuring in the ear of an operator of a vehicle or heavy duty machine may be useful in several applications such as noise assessment, adaptive communication, headphone-less binaural sound reproduction, or active noise control. However, wearing in-ear microphones is not very comfortable for the operator. Therefore, we investigated the possibility to replace these microphones by virtual microphones based on an array of microphones optimally located in an enclosure. Several modifications to LMS based algorithms for sparse MISO system identification were proposed and adopted at the one hand to eliminate as many of the filter taps as possible to reduce the computational load, and at the other hand eliminate as many microphones as possible to reduce hardware costs. Identification methods are evaluated through the simulations based on the experimental measurements performed in the real size driver's cabin constructed for the purpose of active noise control (ANC). Forty six microphones were initially positioned at the cabin roof, while virtual location is chosen to be ear of the seated observer. It is shown that these algorithms indeed reduce the set of microphones, still providing acceptable fitting performance when compared to the reference NLMS technique where all available microphones are used. Sensitivity of converged coefficients to natural head movements is evaluated for the head rotations of 20, 45 and 60 degrees in horizontal plane.