基于深度学习的自主识别智能语音手杖设计  

Design of intelligent speech cane for autonomous recognition based on deep learning

作　　者：王岩[1] 田会峰[1] WANG Yan;TIAN Huifeng(College of Electrical and Information Engineering,Jiangsu University of Science and Technology,Zhangjiagang 215600,China)

机构地区：[1]江苏科技大学电气与信息工程学院,江苏张家港215600

出　　处：《电子设计工程》2022年第17期160-164,169,共6页Electronic Design Engineering

摘　　要：针对老年人与残障人士的出行问题,设计了一款基于深度学习的智能手杖。系统以树莓派为控制核心,驱动树莓派专用的CSI接口并采用Frp内网穿透进行视频流的传输,输入网址即可远程监控。通过按键控制录音与拍摄,分别实现了智能语音交互以及照片自动存云。利用超声波测距模块HC-SR04自动测量系统前方障碍物距离,通过扩音器进行语音报警,通过读取MPU6050模块和GPS模块返回的数据,系统能够对人体姿态与所处位置进行检测。客户端利用TensorFlow Lite深度学习模型实现物体检测识别,并将识别结果展示在显示屏中。设备调试结果表明,系统可并行处理各任务,运行时间快且精度较高。Aiming at the travel problem of the elderly and the disabled,an intelligent walking stick based on deep learning is designed.The system takes Raspberry Pi as the control core,drives the special CSI interface of Raspberry Pi,and uses the Frp intranet penetration to transmit the video stream.The remote monitoring can be realized by inputting the website address.Through the key control recording and shooting,the intelligent voice interaction and automatic cloud storage of photos are realized respectively.The ultrasonic ranging module HC-SR04 is used to automatically measure the distance of obstacles in front of the system.The voice alarm is given through the loudspeaker.The system can detect the human posture and position by reading the data returned by MPU6050 module and GPS module.In the client,TensorFlow Lite deep learning model is used to realize object detection and recognition,and the recognition results are displayed on the display screen.The debugging results of the device show that the system can process all tasks in parallel with fast running time and high precision.

关 键 词：智能手杖 树莓派 云服务器 TensorFlow Lite 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]