面向室内服务的中文语音指令深层信息解析系统  

An indoor service system for parsing of deep information from Chinese voice commands

作　　者：孔令富[1] 高胜男[1] 吴培良[1] 

机构地区：[1]燕山大学信息科学与工程学院,秦皇岛066004

出　　处：《高技术通讯》2014年第11期1101-1107,共7页Chinese High Technology Letters

基　　金：国家自然科学基金(61305113);河北省自然科学基金(F2012203199)资助项目

摘　　要：针对室内服务机器人的人机交互问题,对中文语音指令进行了深入研究,提出了一种基于概率/神经网络混合模型的深层信息解析系统。该系统由指令解析模块和深层信息提取模块组成,前者基于概率模型解析语音指令的有效信息,后者依据家庭环境神经网络模型,将有效信息中的服务对象或目标对象作为已知条件提取指令深层信息,旨在将指令所蕴含的深层信息显性化。构建了一般家庭条件下的实验环境进行了仿真实验,仿真数据验证了指令解析模块和深层信息提取模块的可行性;选取两类典型结构的中文语音指令,在该系统上进行深层信息解析实验,提取了准确的有效信息和深层信息。For realizing the natural human-computer mands was conducted, and a deep information interaction of indoor service robots, a study on Chinese voice comparsing system based on a probability/neural network hybrid model was proposed. This system is composed of a command parsing module and a deep information extraction module, and the former parses out the effective information of voice commands based on the probability model, while the lat- er extracts the deep information, the service object or the target object as a condition known, based on the neural network model of family environment, for making the deep information contained in commands explicit. The experi- mental environment in general family environment was built for simulation experiment, and the simulation data veri- fied the feasibility of the command parsing module and the deep information extraction module. Two kinds of Chi- nese voice commands with typitcal structure were chosen for a deep parising experiment on the system, and the ac- curate effective information and the deep information were extracted.

关 键 词：概率/神经网络混合模型 指令解析模块 有效信息 深层信息提取模块 深层信息 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]