扫频激励下的DPOAE的局部加权最小二乘估计算法

Locally Weighted Least Squares Estimation of DPOAE Evoked by Continuously Sweeping Primaries

畸变产物耳声发射(DPOAE)信号可以用于听力损失的诊断,具有重要的临床应用价值。但是,大频率范围内进行DPOAE测量时,需要耗费大量的时间。采用扫频信号作为诱发信号,为记录大频率范围内的DPOAE测量数据提供了一种快速而有效的手段。本文在已有的最小二乘拟合(LSF)算法的基础上,给出一种估计扫频激励下的DPOAE信号的局部加权最小二乘估计(LWLSE)算法,从扫频激励下的DPOAE记录信号中提取方差根音2f1-f2频率处的DPOAE信号。仿真结果表明,该方法相对于常用的LSF方法,对DPOAE信号和初始刺激信号的估计更准确,并且估计结果具有更小的估计方差,有利于提取更加准确的DPOAE精细结构。

Distortion product otoacoustic emission （DPOAE） signal can be used for diagnosis of hearing loss so that it has an important clinical value. Continuously using sweeping primaries to measure DPOAE provides an efficient tool to record DPOAE data rapidly when DPOAE is measured in a large frequency range. In this paper, locally weighted least squares estimation （LWLSE） of 2fl-f2 DPOAE is presented based on least-squares-fit（LSF） algorithm, in which DPOAE is evoked by continuously sweeping tones. In our study, we used a weighted error function as the loss function and the weighting matrixes in the local sense to obtain a smaller estimated variance. Firstly, ordinary least squares estimation of the DPOAE parameters was obtained. Then the error vectors were grouped and the different local weighting matrixes were calculated in each group. And finally, the parameters of the DPOAE signal were esti- mated based on least squares estimation principle using the local weighting matrixes. The simulation results showed that the estimate variance and fluctuation errors were reduced, so the method estimates DPOAE and stimuli more ac- curately and stably, which facilitates extraction of clearer DPOAE fine structure.