基于智能手机传感器的行为检测研究与应用

发布时间：2018-08-02 09:33

【摘要】：移动终端技术、可穿戴式技术、移动互联网技术、无线传感器技术、嵌入式技术等领域的快速进步及相互结合,推动了智能手机传感器的飞跃发展。智能手传感器将虚拟世界与现实世界紧密地连接起来,改变了人类与环境的交互方式,使虚拟世界信息可以便捷地、有效地表达为现实世界信息。目前,由于可穿戴式设备的快速兴起,智能手机传感器已经广泛应用于用户行为识别、用户行为分析、导航、大型竞速游戏等领域。用户行为识别泛指通过传感器设备收集与用户行为密切相关的大量信息,经过处理后对用户行为进行分析和推理,为用户提供更好的智能服务。本文所研究的用户行为识别主要有两个:驾驶行为检测和行人安全行走检测。对于驾驶行为检测,本文提出了基于加速度传感器的驾驶行为检测方法。由驾驶事件和路面凹凸状况引起的交通事故对司机的驾驶行为来说是非常关键的问题,相关研究人员已经广泛地研究了交通事故的原因和对策。尽管已经提出几种方法来解决这些问题,但是大部分方法都需要很高的计算成本或者是固定额外的硬件设备。为了应对这些挑战,我们设计了一个基于智能手机加速度传感器的系统,HealthDriving,用来检测司机的驾驶事件和路面状况。更具体地说,首先从智能手机的加速度传感器收集加速度数据,然后采用所设计的加速度重定向校准算法将所获得的加速度传感器数据转换为汽车的加速度数据,最后利用HealthDriving来检测司机的驾驶事件和路面凹凸状况。同时,为了评估由司机驾驶行为而引起进攻性程度,采用ISO 2631人体暴露的振动程度标准设计了一个有效的评分机制,对司机的驾驶安全体验进行评分,得分越高,说明司机的驾驶行为越安全,攻击性越低。反之,司机存在严重的危险驾驶行为。大量的评估表明,HealthDriving可以成功运作在普通的智能手机上,并且与其他方法相比,具有较低的计算成本,验证了本方案的可行性和有效性。对于行人的安全行走行为,本文提出了基于加速度传感器和摄像头的行人安全行走检测方法。最近几年,行人在走路时使用智能手机进行阅读或者娱乐活动已经越来越受欢迎,低头看手机行走使行人的注意力集中在手机屏幕上,而忽略了周围环境的不安全。为了避免行人绊倒、跌倒,甚至与其他行人碰撞,我们设计了WalkWell,一个基于智能手机的安全行走检测系统,确保行人在使用手机时的安全。首先使用重力传感器和加速度传感器估计行人的移动速度,然后激活前置摄像头,基于OpenCV4Android检测人脸和眼睛,并通过眼睛灰度图分析瞳孔的运动姿态,说明行人是否正在看手机屏幕。如果看手机屏幕的时间达到了所设的阈值,WalkWell会通过手机振动的方式提醒行人注意安全。我们将Walk Well实现在了Android手机上,而且评估了实验的精确率,实验结果表明,WalkWell可以预防行人走路时长时间看手机屏幕的潜在危险。
[Abstract]:Mobile terminal technology, wearable technology, mobile Internet technology, wireless sensor technology, embedded technology and other fields of rapid progress and integration, promote the rapid development of smart phone sensors. Intelligent hand sensors connect the virtual world to the real world, change the interaction between human and the environment, and make the virtual. The proposed world information can be easily and effectively expressed as the real world information. At present, because of the rapid rise of wearable devices, smart phone sensors have been widely used in the fields of user behavior recognition, user behavior analysis, navigation, large race speed games and so on. In this paper, there are two main types of user behavior identification: driving behavior detection and pedestrian safety walking detection. In this paper, the driving behavior detection based on acceleration sensor is proposed, and the driving behavior detection based on acceleration sensor is proposed. Methods. Traffic accidents caused by driving events and road bump conditions are a key problem for drivers' driving behavior. The researchers have extensively studied the causes and Countermeasures of traffic accidents. Although several methods have been proposed to solve these problems, most of the methods require high computational cost or high cost. To cope with these challenges, we designed a system based on smart phone acceleration sensors, HealthDriving, to detect drivers' driving events and road conditions. More specifically, we first collect acceleration data from the acceleration sensors of the smartphone, and then use the designed acceleration. The degree redirection calibration algorithm converts the obtained acceleration sensor data to the acceleration data of the car. Finally, HealthDriving is used to detect driver's driving events and road bump conditions. At the same time, in order to evaluate the offensive degree caused by driver's driving behavior, a standard of vibration degree of ISO 2631 human exposure is designed. The higher the driver's driving safety experience, the higher the score, the higher the score, the safer driving, the lower the aggressiveness. On the other hand, the driver has serious dangerous driving behavior. A large number of evaluations show that HealthDriving can be successfully operated on a normal smartphone and compared with other methods. Low computing cost proves the feasibility and effectiveness of this scheme. For pedestrian safe walking, a pedestrian safety walking detection method based on acceleration sensor and camera is proposed. In the last few years, pedestrians have been getting more and more popular with their smartphones for reading or entertainment during walking. In order to avoid pedestrians tripping, falling, and even colliding with other pedestrians, we designed WalkWell, a smart mobile detection system based on smart phones, to ensure the safety of pedestrians in the use of a mobile phone. First, the use of gravity sensors and the use of a gravity sensor. The acceleration sensor estimates the moving speed of the pedestrian, then activates the front camera, detects the face and eyes based on the OpenCV4Android, and analyzes the movement of the pupil through the eye grayscale, indicating whether the pedestrians are looking at the mobile screen. If the time of watching the phone screen is reached, the WalkWell will pass through the vibration of the cell phone. We remind pedestrians to pay attention to safety. We implemented Walk Well on a Android phone and evaluated the accuracy of the experiment. The experimental results showed that WalkWell could prevent the potential danger of a long time looking at a mobile phone screen when pedestrians walk.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP212

【参考文献】