机械臂语音控制系统研究

发布时间：2018-04-09 16:35

本文选题：机器听觉　切入点：语音降噪　出处：《西安科技大学》2017年硕士论文

【摘要】：具备各种功能的机器人在现代化社会的各行各业中都得到了广泛应用。在对机器人的使用和控制中,由于传统的控制方式较为单一且操作不便,使得机器人的智能化发展受到限制。因此,如何提高机器人的人机交互能力,找到一种更加便捷的人机交互方式是机器人智能化研究的重要分支。语言是人类交流最主要的方式,若以语音指令去控制机器人使其完成相应的动作将会使人机交互更加便捷高效。本文以机械臂语音控制系统在物流分拣中的应用为研究背景,在分析机械臂软硬件平台基础上对机械臂语音控制系统进行研究。针对噪声干扰影响语音识别准确率的问题,提出一种改进谱减法的语音指令去噪算法,以此来提高语音指令在机械臂语音控制系统中的识别准确率。本文主要对以下几个方面进行研究:首先,在对六自由度MOTOMAN机械臂软硬件分析的基础上,建立基于机器听觉的多路径下物流分拣系统的模型结构,并根据物流分拣系统的不同功能要求,设计各部分的子系统,在控制方式上对其进行扩展。最终确定了通过调用Motocom32动态链接库来实现机械臂对不同类型货物抓取分拣的程序开发思路。其次,针对噪声影响语音识别准确率的问题,提出一种改进谱减法的语音指令去噪算法。该算法根据多窗谱估计理论对语音信号的功率谱进行估计,通过对功率谱密度平滑处理来改进原有的算法。仿真结果验证了改进算法能够更有效的分离出纯净语音,得到信噪比更高的语音指令信号。为克服传统隐马尔科夫模型分类能力弱的缺点,本文将隐马尔柯夫模型和神经网络模型融合,用混合模型来识别去噪后的语音指令。仿真结果表明,混合模型的识别率更高。最后,将本文所提算法应用到机械臂语音控制系统中,并对物流分拣模拟系统进行搭建及运行测试。结果表明,本文所设计开发的相应软件可以实现语音指令控制机械臂对不同类型货物的抓取和分拣,验证了本文所提算法可以应用到机械臂语音控制系统中。
[Abstract]:Robots with various functions have been widely used in various industries in modern society.In the use and control of the robot, the traditional control mode is simple and the operation is not convenient, so the development of the robot intelligence is restricted.Therefore, how to improve the human-computer interaction ability of robots and find a more convenient way of human-computer interaction is an important branch of intelligent robot research.Language is the most important way of human communication. If the robot is controlled by voice instruction to complete the corresponding action, it will make the human-computer interaction more convenient and efficient.Based on the research background of the application of the robot arm voice control system in the logistics sorting, this paper studies the robot arm voice control system based on the analysis of the hardware and software platform of the manipulator.Aiming at the problem that noise interference affects the accuracy of speech recognition, an improved speech instruction denoising algorithm based on spectral subtraction is proposed to improve the accuracy of speech instruction recognition in the robot arm speech control system.This paper mainly studies the following aspects: firstly, based on the software and hardware analysis of 6-DOF MOTOMAN manipulator, the model structure of multi-path logistics sorting system based on machine hearing is established.According to the different functional requirements of the logistics sorting system, the subsystem of each part is designed, and the control mode is extended.Finally, the program development idea of grasping and sorting different types of goods by using Motocom32 dynamic link library is determined.Secondly, aiming at the problem that noise affects the accuracy of speech recognition, an improved spectral subtraction algorithm for speech instruction denoising is proposed.Based on the theory of multi-window spectrum estimation, the power spectrum of speech signal is estimated, and the original algorithm is improved by smoothing the power spectrum density.The simulation results show that the improved algorithm can separate the pure speech more effectively and get the higher signal-to-noise ratio (SNR) of the speech instruction signal.In order to overcome the weakness of traditional Hidden Markov Model (hmm), this paper combines the Hidden Markov Model (hmm) and the Neural Network (NN) model to recognize the de-noised speech instruction.The simulation results show that the recognition rate of the hybrid model is higher.Finally, the algorithm proposed in this paper is applied to the robot arm voice control system, and the logistics sorting simulation system is built and tested.The results show that the corresponding software designed in this paper can be used to control the grab and sorting of different types of goods by the robot arm. It is verified that the algorithm proposed in this paper can be applied to the speech control system of the robot arm.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP241;TN912.3

【参考文献】