脑干听觉诱发电位的子波变换提取技术

概要地论述了诱发电位的主要特点及检测技术，以脑干听觉诱发电位为例，根据其时，频特性，把子波变换技术应用到ＢＡＥＰ的提取中，提出了ＢＡＥＰ的子波变换技术应用到ＢＡＥＰ的提取中，提出了ＢＡＥＰ的子波变换去噪算法；根据ＢＡＥＰ的信噪比较低的实际情况，提出了结合平均技术的改进算法，通过与传统的平均法比较，该方法可以大大减少检测时间和获得较高的信噪比及满意的波形。

The Initial Parameters Estimation of Hidden Markov Model in Speech Recognition

The choice of initial parameters of hidden Markov model (HMM) is very important and much difficult, too. In general, there are two methods in solutions. One is to set the initial parameters of HMM to mean values or nonzero random values, the other method is segmental K means procedure. But the above methods may cause the Baum Welch algorithm converging to a partial optimum solution, or increase the amount of computation. In order to solve the above defaults, the authors firstly analyzed the relations between observation sequences and parameters of HMM on certain conditions. Then, the practicable estimation of initial parameters of HMM is given.