To estimate the underlying source waveform from the observations is always an important issue in signal processing field. In this paper, a simple method was presented to estimate the initial value and the temporal autocorrelation function instead of the direct estimation of the underlying source. A set of nonlinear equations was constructed based on temporal autocorrelation function of the underlying source signal and solved with the large-scale numerical method. The initial guess and the temporal autocorrelation function were estimated using the wavelet transform modulus maxima method and the estimation of the original signal was obtained after the convergence of the iteration. This method was applied to extract the evoked potential signal with a single channel and single trial. The correlation coefficient between the estimated signal and the source was approximately 0.93 when signal-to-noise ratio was 0dB, both amplitude and latency of the evoked potential signal were accurately estimated. Experimental results showed that the performance of the proposed algorithm was affected by the accuracy of the initial estimated values for both evoked potentials and the temporal autocorrelation function, especially for the latter.%从观测数据中估计源信号的波形一直是信号处理领域中的一个重要问题.利用源信号的时间自相关函数构造一个非线性方程组,借助大规模数值计算方法,把原来直接估计源信号这样一个较困难的问题,转化为估计迭代初始值与源信号时间自相关函数的问题.使用小波模极大值法,估计迭代初始值与源信号时间自相关函数,再对方程组迭代求解,收敛后的结果即为对源信号的估计.将该方法应用于诱发电位的单通道、单次提取实验,在信噪比为0 dB时,提取结果与源信号的相关系数约为0.93,对幅度与潜伏期的估计都比较准确.实验结果表明,算法性能受迭代初始值估计精度与源信号时间自相关函数估计精度的影响,受前者影响较小,而受后者的影响相对较大.
展开▼
