【作者】 李春晓；

【导师】 赵旦峰；

【作者基本信息】 哈尔滨工程大学 ， 通信与信息系统， 2006， 硕士

【摘要】 本课题是研究短波通信中莫尔斯信号的自动检测和识别。莫尔斯信号是短波信号中简单、实用的一种通信制式，目前仍被大量的采用着，尤其在军事通信领域，更发挥着不可替代的作用。长期以来，莫尔斯信号的接收译码都靠人工完成，大量人力每天都在做简单的重复性工作。为了解放人力减轻工作强度，本课题研究并实现了莫尔斯信号的自动检测和识别抄收技术，以替代人工手抄的工作过程。 语音信号的频率范围介于100Hz至3400Hz之间，语音识别技术就是利用数字信号处理方法提取该频段范围信号特征，再依据一定的判别准则判定信号的内容。莫尔斯信号是单音频等幅信号，频率一般出现在1000Hz左右，位于语音频段内，并且信号表达式简单，故可利用语音信号处理方法去处理莫尔斯信号，达到检测和识别的目的。 提取语音参数之前，有一些经常使用的、共同的短时分析技术必须预先进行，如语音信号的数字化、语音信号的端点检测、预加重、加窗和分帧等，这些是不可忽视的语音信号分析的关键技术。 语音信号的参数有时域和频域之分。本文利用时域参数中的短时能量和短时过零率作为莫尔斯信号端点检测参数取得了很好的效果。 噪声是短波信号中不可避免的成分，所以选用合适的滤波方法，得到干净的信号质量是必需环节。功率谱法是一种广泛用于去除加性噪声的技术。应用于短波信号中能去除大部分背景噪声，提高信号质量。 当今语音识别技术的主流算法，主要有隐马尔可夫模型(HMM)方法和矢量量化(VQ)方法。矢量量化的算法所需的模型训练数据，训练与识别时间，工作存储空间都很小。对于词汇量很小的莫尔斯信号非常适用。 莫尔斯信号采用的速率很广，从60码／分到200码／分不等，采用错误的速率去识别未知信号，必然导致错误的结果。本文根据莫尔斯信号特点，应用统计平均值算法，动态估计当前信号的速率，从而得到正确的识别结果。 本文通过对语音识别技术的深入研究，结合莫尔斯报文的特点，实现了更多还原

【Abstract】 This thesis researches the auto-detection and recognize of morse signal which is simple and applied in the shortwave communication. Today it still was applied widely, especially in the military area stand the place where it cannot be instead. In long term, the work of decoding the morse’s signal were done by manual, a lot of people done the simple and repeatable work day by day. To release the manpower, the paper researchs and achieve the auto-detect and recognize the morse signal so that insteading of the manual work.The frequency range of the speech signal is from 100Hz to 3400Hz. The speech recognize technology do pick up this range’s characteristics with digital signal process, then according to the certain deterninal rule determinate the signal’s content. The morse’s signal is singl frequency and fixed intensity. It’s frequency generally appear near 1000Hz which is in the speech frequency’s range. Because the signal’s expression is simple, it is feasible to process the morse’s signal with digital signal process means and to achieve the detection and recognize.Before pick up the speech’s characteristics, it is must that take the common use short-time analyzed technology on the speech signal, for example the analog to digital convertion, the end-point detection, the pre-emphasis, taking windows and so on. These are the key technics in the speech analysis.The parameters of Speech signal include tow parts: the time-range and the frequency-range. This thesis use the short-time energy and the short-time crossing zero radio to detect the morse signal’s end-point and achieve the good effect.Noise is disavoid in shortwave communication. So selecting fit filter to acquier cleanly signal is very important. Powertram is used to filter the plus noise. In the shortwave signal this technology can filter a majority of the background noise and improve the signal quality.更多还原