传统超声回波时延估计算法是在高斯噪声背景下展开研究的, 而实际工况中超声回波不仅含有高斯噪声, 还含有脉冲冲击噪声(α稳定分布噪声)等, 这导致传统算法失效。为了解决上述问题, 本文提出了一种针对混合噪声特别是包含噪声背景下的超声回波时延估计算法: 归一化循环相关时延估计算法。首先, 对归一化循环相关算法理论进行了简要的介绍。接着, 对归一化循环相关时延估计算法进行了理论推导分析。然后, 结合仿真分析, 在相同α混合噪声情况下对传统循环相关和归一化循环相关时延估计进行比较。最后, 在不同信噪比下, 对归一化循环相关时延估计算法的估计性能进行了分析。通过对比实验发现, 在噪声特征指数趋于1时, 循环相关算法已不能估计出时延, 而归一化循环相关算法的误差仍能保持在0.4 μs; 且在-10dB信噪比下, 归一化循环相关算法时延估计也能保持在10 μs误差范围内。本文所提归一化循环相关算法在混合噪声特别是包含α噪声情况下能够对超声回波时延进行精确估计, 具有传统算法所不能比拟的优势。

Traditional ultrasonic echo time-delay estimation algorithm is studied in the background of Gauss Noise, while in actual working condition ultrasonic echo includes not only Gauss Noise but also impulsive noise (α noise with stable distribution), which will invalidate traditional algorithms. In actual application, in order to solve the above problem, a new ultrasonic echo time-delay estimation method under mixed noise, especially the background containing α noises, was proposed, called time delay estimation algorithm with normalized cyclic correlation. Firstly, a brief introduction to theory of such normalized cyclic correlation algorithm was given and a theoretical derivation analysis on such algorithm was conducted.Then,combined with simulation analysis,the performance of proposed algorithm and traditional cyclic cross-correlation algorithm were compared under the same condition. At last, the estimation performance of such normalized cyclic correlation time delay estimation algorithm was analyzed under different singal-noise ratios. Contrast test indicates that when α approaches to 1, the cyclic cross-correlation algorithm is unable to estimate the time delay, while the normalized cyclic cross-correlation algorithm still maintains the deviation at 0.4 μs; and under -10dB SNR, the time delay estimated by normalized cyclic cross-correlation algorithm is also maintained within 10 μs. The normalized cyclic cross-correlation algorithm proposed in the article is able to make accurate estimation of time delay of ultrasonic echo under the background of mixed noise, especially the one contains α noise, which is an incomparable advantage that traditional algorithm is unable to counteract.