基于连续隐马尔可夫模型的协作频谱检测

摘要

机器学习是当前人工智能的主要研究方向,连续隐马尔可夫模型( Continuous Hidden MarKov Model,CHMM)作为机器学习方法的一种被广泛应用于故障诊断、图像处理、生命科学等领域。研究表明,在信道占用和空闲状态下采样得到的能量值满足不同的高斯分布,故可采用机器学习方法通过模式识别进行频谱感知；同时为了克服离散隐马尔可夫模型( Discrete Hidden MarKov Model,DHMM)在处理连续信号矢量量化过程中产生的信息失真问题,文中将CHMM引入多用户协作频谱检测技术,分别根据信道占用和信道空闲时采集到的能量值来训练CHMM模型建立CHMM1-CHMMn ,多个次用户分别将当前采集到的信道的能量值作为待测矩阵同CHMM1-CHMMn进行模式识别,根据识别结果判定当前信道是占用还是空闲。仿真结果表明,该方法在频谱感知方面具有较高的准确性。%Machine learning is the main research direction of artificial intelligence now,Continuous Hidden Markov Model ( CHMM) is widely used in the fields of fault diagnosis,image processing,life science and others as a machine learning method. Research has shown that the collected energy values in channel occupied status and channel idle status meet different Gaussian distribution,so spectrum sensing can be carried out with machine learning method by pattern recognition. At the same time,in order to overcome the information distortion problem caused by DHMM when processing vector quantization, use CHMM in multi-user cooperative spectrum detection, training CHMM models to build CHMM1-CHMMn based on the collected energy values in channel occupied status and channel idle status respec-tively,multiple secondary users treat the collected energy values as the testing matrix to match with CHMM1-CHMMn ,to judge the chan-nels whether occupied or idle based on the match result. The simulation results show that this method has high accuracy in spectrum sens-ing.