【摘要】 针对现有步态识别研究中步态识别率低、算法单一等问题,提出了一种基于惯性运动传感器的步态识别方法。首先,该方法结合动态时间规整与人工神经网络,通过前者提取固定长度的步态特征,并设置成本函数的阈值来判别后者的正负输入,提取得到波形特征。其次,运用列文伯格-马夸尔特算法改进标准BP神经网络,最终完成步态识别。实验证明所提的改进步态识别方法将平均步态识别率和相等错误率维持在91. 5%和9. 1%,较好地提高了步态识别的准确率。因此,该方法可作为高级认证方法的补充,以增加个人信息的隐私和安全性,适合实验室仪器安全管理应用。更多还原

【Abstract】 To solve the problems of low gait recognition rate and single algorithm in current gait recognition research,a gait recognition method based on inertial motion sensor is proposed. First,this method combines dynamic time warping with artificial neural network,extracts fixed-length gait features by the former and sets the threshold value of cost function to discriminate the positive and negative inputs of the latter,and extracts the waveform features. Second,the Levenberg-Marquardt algorithm is used to improve the standard BP neural network. Finally,the gait recognition is completed. Experimental results show that the improved gait recognition method maintains the average gait recognition rate and the equal error rate at 91. 5% and 9. 1%,which improves the accuracy of gait recognition. Therefore,the proposed method can be used as a supplement to the high-level authentication method to increase the privacy and security of personal information. And it is also very suitable to the application in the secure management of the apparatus in laboratory.更多还原