Direction of Arrival Estimation for Wideband Sources Based on Time-Frequency Signal Representation
LI Zhi-shun · Acta Simulata Systematica Sinica · 2004
A novel wideband Direction of Arrival estimation method is proposed by exploiting the signals different time-frequency signatures. The data vector model is constructed in time-frequency domain by computing the cross Wigner-Ville distribution of the output of the reference sensor and those of other sensors. Accordingly the time-frequency correlation matrix for each source is formed by choosing the t-f points residing on its own auto-term region, separately. Then the traditional subspace based method is applied on this t-f correlation matrix to obtain the DOA estimates of multiple wideband sources. In contrast to the traditional method, the proposed method combines time-frequency analysis into array processing, and the DOA estimating performance is improved due to localizing the source energy while spreading the noise power in the time-frequency domain. Furthermore, the method is robust against noise and has a low computation load in constructing the time-frequency correlation matrix compared to the traditional t-f method. The computer simulation verified the efficiency of the new method.