Acoustic Feedback Reduction Based on FIR and IIR Adaptive Filters in ITE Digital Hearing Aids
Acoustic feedback phenomenon can deteriorate the performance of digital hearing aids performance working at high gains, causing instability and speech degradation. In order to restore a stable situation, it is needed a feedback reduction system based on adaptive algorithms. Depending on the adaptive flter impulse response, a feedback reduction algorithm based on a finite impulse response (FIR) or an infnite impulse response (IIR) adaptive filter can be used. The commonly used adaptive algorithm based on FIR filters is the least-mean square (LMS) algorithm, but it becomes unstable if some conditions are not fulfilled. In order to avoid this problem, a feedback reduction subsystem based on IIR filters is proposed. In this case, the IIR filter is adapted with a modified version of the LMS algorithm for this kind of filters, which is called the IIR-LMS algorithm. These two adaptive algorithms are tested in the In-The-Ear (ITE) hearing aid category, which the category that most suffer from acoustic feedback. In order to determine which feedback reduction technique (FIR or IIR) is the most suitable for the ITE hearing aid, the added stable gain (ASG) value over its limit gain of performance is obtained. The ASG value is achieved as a tradeo between the speech quality (subjective parameter) and the Segmented Signal-to-Noise Ratio (objective parameter). The results show how the digital hearing aid working with an acoustic feedback reduction subsystem based on IIR filters adapted with the IIR-LMS algorithm achieves an increase of 8 dB in its ASG.
A.Martinez-Leira R.Vicen-Bueno R.Gil-Pita M.Rosa-Zurera
Dimetronic Signals San Fernando Business Park (Hungria Building) Avda.Castilla 2, 28830, San Fernand Signal Theory and Communications Department Escuela Politfiecnica Superior, Universidad de Alcalfia
国际会议
2008 International Conference on Audio，Language and Image Processing(2008国际声音、语言、图像过程大会)
镇江
英文
1442-1448
2008-07-07（万方平台首次上网日期，不代表论文的发表时间）