为了提高气体传感器漂移故障检测的灵敏度,采用了一种基于多分辨分析理论的在线小波分解算法(OLMS) . 与传统的小波分解算法不同的是,OLMS 算法具有较好的实时性,适于在线使用. 采用该算法可以将传感器输出信号中发生在不同的时间尺度上的“事件”分离开,因此能够将变化微弱的漂移特征从信号噪声和其他干扰中有效地检测出来. 此外,对于采集序列中常见的“野点”数据,提出一种改进的中位值滤波算法(MMF) 予以消除. 仿真计算结果证明了该算法的有效性.关键词: 在线小波变换;传感器漂移;中位值滤波Abstract : In order to improve the sensitivity of drift detection for gas sensors , an on2line wavelet de2 composition algorithm (OLMS) is applied. Unlike the t raditional wavelet decomposition algorithm , OLMS is adaptively used on2line. Applying this algorithm , the sensor signal can be decomposed into multi scales. This makes the drift features separated sensitively f rom the noises or other event s. More2 over , a modified median filtering method is proposed to eliminate the outliers in the collected data suc2 cessfully. Detail des cription of the algorithm is given and it s superior performances are illust rated by simulation.Keywords : on2line w avelet decom position ; sensor drif t ; median f il ter