脉冲噪声下多伯努利滤波的单声矢量DOA跟踪

针对单声矢量传感器(Acoustic vector sensor, AVS)脉冲噪声环境下的多声源波达方向(Direction of arrival, DOA)跟踪问题,利用α稳定分布能更好地建模脉冲噪声的性质,提出α稳定分布下的多伯努利DOA跟踪算法。由于α稳定分布不具有有限协方差,该算法采用分数低阶距(Fractional Lower Order Moment, FLOM)代替协方差矩阵,对FLOM进行特征分解构造噪声子空间,生成FLOM MUSIC空间谱函数作为多伯努利滤波器的伪似然函数,并对其指数加权,改善了传统似然函数的发散和平坦问题,使得粒子的重采样更有效。该算法的优点是不需要预先知道声源个数,利用先验信息和当前量测信息可以直接对当前声源进行跟踪。仿真结果表明,该算法能有效跟踪脉冲噪声环境下单一AVS声源的数目和状态。 

DOA tracking of single acoustic vector sensor by Multi-Bernoulli filter in impulse noise

Aiming at the tracking problem of multi-source Direction of arrival (DOA) in the pulse noise environment of Acoustic vector sensor (AVS), the DOA tracking algorithm based on α stable distribution that can be better models the properties of impulse noise is proposed. Since the α stable distribution does not have finite covariance, the covariance matrix is replaced by the fractional low order moment (FLOM). The noise subspace is constructed by eigendecomposition of FLOM, and FLOM-MUSIC spatial spectrum function is generated as pseudo-likelihood function of Multi-Bernoulli filter. By exponential weighting, the problem of divergence and flatness of the traditional likelihood function is improved, and the resampling of the likelihood function in Gaussian region is more effective. The advantage of this algorithm is that it is not necessary to know the number of sound sources in advance, and the current sources can be directly tracked by using the prior information and the current measurement information. The simulation results show that the algorithm can effectively track the number and state of sources of a single AVS in the impulse noise environment.