采用卷积神经网络的老年人跌倒检测系统设计
|
摘要
为了利用便携式设备准确检测老年人的跌倒状况,针对传统算法中人为设计特征造成的不完备性,构建一种基于卷积神经网络(CNN)的老年人跌倒检测模型.以智能手机内置的三轴传感器作为数据获取源,将采集的人体姿态信息进行滤波、标准化、采样等操作后,输入到所设计的模型中;采用梯度下降和适应性动量优化方法进行多层卷积神经网络训练和优化,获得模型关键参数训练并优化模型关键参数;利用学习到的深层次特征进行样本分类.实验结果表明:所设计的模型对于跌倒检测的准确率明显高于一般的机器学习算法模型,并且在对跌倒和非跌倒的区分检测中,精确率和召回率都保持了较高的稳定水平.
A fall detection model based on convolutional neural network(CNN) was constructed, in order to use portable device to accurately detect the fall of the elderly and to avoid the incompleteness caused by the artificial designed features in traditional algorithms. The three-axis sensor built in the smart phone was used as the data acquisition source; the collected human body posture information was filtered, standardized, and sampled, etc. and then inputted into the designed model. Multi-layer CNN was used to train and optimize the key parameters of the model, combined with gradient descent and adaptive moment estimation optimization methods; the learned deep features were used for samples, classification. The experimental results show that the accuracy of the designed model for fall detection is significantly higher than that of the general machine learning algorithm model. In addition, the evaluation indicators have maintained a high level of stability in the detection of falls and non-falls.
A fall detection model based on convolutional neural network(CNN) was constructed, in order to use portable device to accurately detect the fall of the elderly and to avoid the incompleteness caused by the artificial designed features in traditional algorithms. The three-axis sensor built in the smart phone was used as the data acquisition source; the collected human body posture information was filtered, standardized, and sampled, etc. and then inputted into the designed model. Multi-layer CNN was used to train and optimize the key parameters of the model, combined with gradient descent and adaptive moment estimation optimization methods; the learned deep features were used for samples, classification. The experimental results show that the accuracy of the designed model for fall detection is significantly higher than that of the general machine learning algorithm model. In addition, the evaluation indicators have maintained a high level of stability in the detection of falls and non-falls.
引文
[1]师昉,李福亮,孙思佳,等.中国老年跌倒研究的现状与对策[J].中国康复,2018,33(3):246-248.SHI Fang,LI Fu-liang,SUN Si-jia,et al.The present situation and countermeasures of the study of senile falls in China[J].Chinese Journal of Rehabilitation,2018,33(3):246-248.
[2]陈瑜,姜小鹰,杨美丽,等.养老机构老年人跌倒效能状况及其影响因素分析[J].中国实用护理,2016,32(6):412-415.CHEN Yu,JIANG Xiao-ying,YANG Mei-li,et al.Investigation on the fall efficacy of elderly people in nursing homes and its influencing factors[J].Chinese Journal of Practical Nursing,2016,32(6):412-415.
[3]KHAN S S,HOEY J.Review of fall detection techniques:a data availability perspective[J].Medical Engineering and Physics,2017,1(39):12-22.
[4]GHEN K H,HSU Y W,YANG J J,et al.Enhanced characterization of an accelerometer-based fall detection algorithm using a repository[J].Instrumentation Science and Technology,2016,11(6):382-391.
[5]MADARSHAHIAN R,CAICEDO J M,ZAMBRANAD A.Evaluation of a time reversal method with dynamic time warping matching function for human fall detection using structural vibrations[C]//Model Validation and Uncertainty Quantification,Volume 3,Conference Proceedings of the Society for Experimental Mechanics Series.Florida:Springer,2014:171-176.
[6]CHEN K H,YANG J J,JAW F S.Accelerometerbased fall detection using feature extraction and support vector machine algorithms[J].Instrumentation Science and Technology,2016,44(2):333-342.
[7]LIM D,PARK C,KIM N.Fall-detection algorithm using 3-axis acceleration:combination with simple threshold and hidden markov model[J].Journal of Applied Mathematics,2014,9(17):342-351.
[8]AZIZ O,MUSNGI M,PARK E J,et al.A comparison of accuracy of fall detection algorithms(thresholdbased vs.machine learning)using waist-mounted triaxial accelerometer signals from a comprehensive set of falls and non-fall trials[J].Medical and Biological Engineering and Computing,2017,55(1):45-55.
[9]OZDEMIR A,BARSHAN B.Detecting falls with wearable sensors using machine learning techniques[J].Sensors,2014,14(6):10691-10708.
[10]MONGIA M,KUMAR K,ERRAQABI A,et al.On random weights for texture generation in one layer CNNS[C]//ICASSP IEEE International Conference on Acoustics.New Orleans:IEEE,2017:2207-2211.
[11]XIA Y,NAREN W L,WANG K Q,et al.Detecting atrial fibrillation by deep convolutional neural networks[J].Computers in Biology and Medicine,2018,93(2):84-92.
[12]ABADI M,BARHAM P,CHEN J M,et al.TensorFlow:a system for large-scale machine learning[C]//Proceedings of the12th USENIXSymposium on Operating Systems Design and Implementation.Savannah:USENIX,2016:265-283.
[13]Activity recognition challenge-dataset[EB/OL].[2014-11-17].http://www.opportunity-project.eu/index.html.
[2]陈瑜,姜小鹰,杨美丽,等.养老机构老年人跌倒效能状况及其影响因素分析[J].中国实用护理,2016,32(6):412-415.CHEN Yu,JIANG Xiao-ying,YANG Mei-li,et al.Investigation on the fall efficacy of elderly people in nursing homes and its influencing factors[J].Chinese Journal of Practical Nursing,2016,32(6):412-415.
[3]KHAN S S,HOEY J.Review of fall detection techniques:a data availability perspective[J].Medical Engineering and Physics,2017,1(39):12-22.
[4]GHEN K H,HSU Y W,YANG J J,et al.Enhanced characterization of an accelerometer-based fall detection algorithm using a repository[J].Instrumentation Science and Technology,2016,11(6):382-391.
[5]MADARSHAHIAN R,CAICEDO J M,ZAMBRANAD A.Evaluation of a time reversal method with dynamic time warping matching function for human fall detection using structural vibrations[C]//Model Validation and Uncertainty Quantification,Volume 3,Conference Proceedings of the Society for Experimental Mechanics Series.Florida:Springer,2014:171-176.
[6]CHEN K H,YANG J J,JAW F S.Accelerometerbased fall detection using feature extraction and support vector machine algorithms[J].Instrumentation Science and Technology,2016,44(2):333-342.
[7]LIM D,PARK C,KIM N.Fall-detection algorithm using 3-axis acceleration:combination with simple threshold and hidden markov model[J].Journal of Applied Mathematics,2014,9(17):342-351.
[8]AZIZ O,MUSNGI M,PARK E J,et al.A comparison of accuracy of fall detection algorithms(thresholdbased vs.machine learning)using waist-mounted triaxial accelerometer signals from a comprehensive set of falls and non-fall trials[J].Medical and Biological Engineering and Computing,2017,55(1):45-55.
[9]OZDEMIR A,BARSHAN B.Detecting falls with wearable sensors using machine learning techniques[J].Sensors,2014,14(6):10691-10708.
[10]MONGIA M,KUMAR K,ERRAQABI A,et al.On random weights for texture generation in one layer CNNS[C]//ICASSP IEEE International Conference on Acoustics.New Orleans:IEEE,2017:2207-2211.
[11]XIA Y,NAREN W L,WANG K Q,et al.Detecting atrial fibrillation by deep convolutional neural networks[J].Computers in Biology and Medicine,2018,93(2):84-92.
[12]ABADI M,BARHAM P,CHEN J M,et al.TensorFlow:a system for large-scale machine learning[C]//Proceedings of the12th USENIXSymposium on Operating Systems Design and Implementation.Savannah:USENIX,2016:265-283.
[13]Activity recognition challenge-dataset[EB/OL].[2014-11-17].http://www.opportunity-project.eu/index.html.