基于智能手机的行人位移检测方法研究

发布时间：2019-04-09 09:08

【摘要】：人工智能技术的发展和支持惯性传感器的可穿戴设备的不断普及,促进了行为识别相关研究和应用的快速发展。作为行为识别研究的一个重要分支,行人位移检测在设备定位、节能、医疗看护等方面都具有十分重要的作用。目前研究与应用主要面临两个问题:依赖专用的惯性传感器设备,系统成本较高;传感器设备位置固定,系统灵活性低。与此同时,日渐普及的智能手机已普遍嵌入了加速度计、陀螺仪和磁力计等惯性传感器,通过智能手机检测行人位移显然将具有更加广泛的应用价值。本文提出了一种基于智能手机内嵌陀螺仪的行人位移检测方法。首先,通过使用快速傅立叶变换(Fast Fourier Transform,FFT)方法,有效提取陀螺仪信号的频域特征。其次,根据非约束条件下行人步行的频率特征,建立了一个基于阈值的检测模型,实现了位移的准确检测。最后,在真实场景中设计并实现大量的实验,通过与现有的标准差阈值(Standard Deviation Threshold,STD__TH)方法和短时傅里叶变换(Short Term Fourier Transform,STFT)方法进行比较,验证了本文提出的行人位移检测方法的有效性和可行性。实验结果表明,本文提出方法的识别精度保持在83%以上,最高可达92.66%;而STFT方法和STD_TH方法的精度较差,最高精度可达79.40%和67.80%。综上所述,本文提出的位移检测方法在位置约束、检测精度等方面都展现了一定的优势,将对基于行为识别的应用和研究起到积极的作用。
[Abstract]:The development of artificial intelligence technology and the popularization of wearable devices which support inertial sensors promote the rapid development of research and application related to behavior recognition. As an important branch of behavior recognition, pedestrian displacement detection plays an important role in equipment location, energy saving, medical care and so on. At present, the research and application mainly face two problems: depending on the special inertial sensor equipment, the system cost is high, the sensor equipment is fixed in position, and the system flexibility is low. At the same time, inertial sensors such as accelerometer, gyroscope and magnetometer are widely embedded in the increasingly popular smart phones. It is obvious that detecting pedestrian displacement through smart phones will have a wider application value. In this paper, a pedestrian displacement detection method based on smart phone embedded gyroscope is proposed. Firstly, by using Fast Fourier transform (Fast Fourier Transform,FFT) method, the frequency domain features of gyroscope signal are extracted effectively. Secondly, according to the frequency characteristics of pedestrian walking under unconstrained conditions, a threshold-based detection model is established, and the accurate detection of displacement is realized. Finally, a large number of experiments are designed and implemented in real-world, and compared with the standard difference threshold (Standard Deviation Threshold,STD__TH) method and the short-time Fourier transform (Short Term Fourier Transform,STFT) method. The validity and feasibility of the proposed pedestrian displacement detection method are verified. The experimental results show that the recognition accuracy of the proposed method is above 83% with a maximum of 92.66%, while the accuracy of the STFT method and the STD_TH method is poor, with the highest accuracy of 79.40% and 67.80%. In summary, the displacement detection method presented in this paper shows some advantages in position constraint, detection accuracy and so on, which will play an active role in the application and research of behavior-based recognition.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9

【参考文献】