基于语音识别的机器人控制技术研究

发布时间：2018-04-27 11:38

本文选题：语音识别 + MFCC　；参考：《西南石油大学》2014年硕士论文

【摘要】：随着当今科技的高速发展,语音识别技术被越来越多的人所关注。语音识别技术作为智能机器人研究领域的一个重要分支,其目的就是让机器能够听懂人类的语言,便于人机交流。因此,将语音识别技术应用于机器人控制领域,体现了当今自动化的发展水平。本论文基于语音识别技术,以实现机器人的简单运动控制为目标,完成对特定人孤立词语音信号的识别,重点研究语音信号的特征参数提取算法和语音识别算法。结合语音识别的基础理论知识,对采集到的语音信号进行时域和频率域的分析以及归一化、预加重、加窗分帧、端点检测等预处理变换。以梅尔频率倒谱系数(MFCC)作为语音信号的特征参数序列,并将MFCC与小波变换相结合,获得更能有效表征语音信号特征的WT-MFCC参数。对比目前几种常用语音识别算法的特点,选择动态时间规整(DTW)算法和神经网络算法对语音信号进行识别。对DTW算法和BP神经网络算法进行深入研究,使用MATLAB编写基于DTW和BP神经网络的语音识别程序,对比识别效果并改进识别算法。在LabVIEW中编写语音信号采集程序以及语音识别的上位机界面,在Proteus中搭建下位机直流电机正反转控制电路,通过虚拟串口技术,完成纯软件环境下的语音识别控制系统的仿真。通过上位机语音识别系统对采集到的语音命令进行识别,编写MT-UROBOT机器人的运动控制程序,采用DTD462无线通信模块进行无线通信,实现对MT-UROBOT机器人的语音控制。
[Abstract]:With the rapid development of science and technology, more and more people pay attention to speech recognition technology. As an important branch of intelligent robot research, speech recognition technology aims at enabling machines to understand human language and facilitate human-computer communication. Therefore, the application of speech recognition technology in robot control field reflects the development level of automation. Based on speech recognition technology, this paper aims at the realization of simple motion control of robot, accomplishes the speech signal recognition of isolated words of a specific person, and focuses on the feature parameter extraction algorithm and speech recognition algorithm of speech signal. Combined with the basic theory of speech recognition, the speech signals collected are analyzed in time domain and frequency domain, and the preprocessing transformation, such as normalization, pre-weighting, windowed frame splitting, endpoint detection and so on, is carried out. Using Mel frequency cepstrum coefficient (MFCC) as the characteristic parameter sequence of speech signal and combining MFCC with wavelet transform, the WT-MFCC parameters which can represent the feature of speech signal more effectively are obtained. Compared with the characteristics of several commonly used speech recognition algorithms, dynamic time warping (DTW) algorithm and neural network algorithm are selected to recognize the speech signal. The DTW algorithm and BP neural network algorithm are deeply studied. The speech recognition program based on DTW and BP neural network is compiled by MATLAB. The recognition effect is compared and the recognition algorithm is improved. The speech signal acquisition program and the upper computer interface of speech recognition are written in LabVIEW, and the forward and backward control circuit of DC motor is built in Proteus. The simulation of speech recognition control system under the pure software environment is completed through virtual serial port technology. Through the speech recognition system of the upper computer, the voice command is recognized, the motion control program of the MT-UROBOT robot is compiled, and the wireless communication module of DTD462 is used to realize the speech control of the MT-UROBOT robot.
【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP242;TN912.34

【参考文献】