论文部分内容阅读
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。
Aiming at the man-machine interaction problem of indoor service robots, the Chinese phonetic instruction is deeply studied, and a deep information analysis system based on the hybrid model of probability / neural network is proposed. The system consists of an instruction parsing module and a deep information extraction module. The former analyzes effective information of voice commands based on the probabilistic model. The latter, based on the neural network model of family environment, extracts the service objects or target objects in the effective information as known conditions The message is designed to make the underlying information implicit in the directive explicit. The experimental environment under normal home conditions was constructed and simulated. The simulation data verified the feasibility of the instruction analysis module and the deep information extraction module. Two kinds of typical Chinese phonetic instructions were selected, and deep information analysis experiments were conducted on the system. Extract accurate and valid information and deep information.