Almost perfect sequences based on cyclic difference sets

The perfect sequences are so ideal that all out-of-phase autocorrelation coefficients are zero, and if the sequences are used as the coding modulating signal for the phase-modulated radar, there will be no interference of side lobes theoretically. However, it has been proved that there are no binary perfect sequences of period 4 〈 n ≤ 12100. Hence, the almost perfect sequences with all out-of-phase autocorrelation coefficients as zero except one are of great practice in engineering. Currently, the cyclic difference set is one of most effective tools to analyze the binary sequences with perfect periodic autocorrelation function. In this article, two characteristic formulas corresponding to the autocorrelation and symmetric structure of almost perfect sequences are calculated respectively. All almost perfect sequences with period n, which is a multiple of 4, can be figured out by the two formulas. Several almost perfect sequences with different periods have been hunted by the program based on the two formulas and then applied to the simulation program and practical application for ionospheric sounding.

基于单边自相关线性预测噪声中汉语语音识别

对含噪语音在自相关域上进行处理 ,以其自相关函数值为参数进行端点检测 ,以基于单边自相关序列的LPC倒谱系数作为语音的特征参数进行语音识别 ,实验表明 :这种方法较好地消除了噪声对语音信号的干扰 ,并获得了较高的识别率。在信号的信噪比低而自相关性又强时 ,此法能体现出不同一般的优势 ,为实际应用提供了可能。

噪声环境下语音识别方法研究

研究了 6种噪声背景下与说话人有关的孤立词语音识别方法。它们是 :线性预测误差法 ,单边自相关线性预测法 ,语音前端声学处理法 ,正则相关分析的谱变换补偿方法 ,特征综合法和同模极点增加法。实验结果表明 ,这 6种方法都有效地提高了噪声环境中语音识别率 ,其中较好的方法在强噪声环境中 (信噪比为 0dB)的语音识别率达到 80 %以上 。

一种基于相对自相关序列的语音端点检测法

在研究单边自相关序列(One-sided autocorrelation,OSA)和相对自相关序列(Relative autocorrelation sequences,RAS)的基础上,提出了一种基于相对自相关序列的语音信号的端点检测算法。该方法利用相对自相关算法能够消除噪声的原理,以语音信号相对自相关序列短时平均幅度代替双门限比较法中的语音信号短时平均幅度,以语音信号短时平均幅度代替语音信号的短时平均过门限率,实验表明,在低信噪比下检测精度要高于传统的双门限比较法。