提出了一种基于加速度计、陀螺仪、气压计输出时域特征的高精度、高实时性的人体行为模式识别算法.该算法选取多传感器输出的时域特征值作为唯一特征量,通过特征提取运算实现行为的实时识别.通过在实验室自主研发的软硬件平台上进行测试,在识别时间缩短到2s一次的条件下,对于8种人体日常行为模式和4种摔倒模式的平均识别率可达到94%以上.该算法对于现有算法实时精度有明显提高,且拓展了模式识别的种类,在可穿戴智能终端领域具有很好的应用前景.%An algorithm of activity pattern recognition based on the time-domainfeatures of accelerometer, gyroscope and barometerwasproposed forhigh-accuracyreal-time human activity pattern recognition. The time-domainfeatureobtained from multi-sensoris selected as the only feature parameter,andtheactivity recognitionis realized through feature extraction operation. The data testson the independent hardware and software platformsin the laboratory indicate thatthe averagerecognitionrate,for thereal-time activitywith8 kinds of daily activities and 4 kinds of falling downactivities,is above 94%withextremely shortrecognition timeof about2s each time.Thetestresult proves that this algorithmnot only significantly improves the real-time precisionofthepresentalgorithms,but alsoexpands the typesof activity recognition, whichhas great application prospectin wearable intelligentterminal areas.
展开▼