Chinese speech recognition and task analysis of aldebaran Nao robot

Abstract

It is one of the important goals for human-computer interaction to make the robot understand the Chinese instruction. The paper presents a scheme for Chinese instruction processing. After the user's speech instruction is obtained, the first thing to do is to conduct a voice preprocessing and Voice Activity Detection (VAD), and then the speech recognition by means of Baidu Cloud Speech Recognition technology. For recognized sentences, Chinese segmentation is carried out by using the Chinese segmentation algorithm combining dictionary and statistics. Then the professional keyword database is utilized to extract the key words, with the aim of attaining the verbs to perform the task and the nouns related to the places and objects in the task, so that the robot can receive the task semantic analysis. The method proves to achieve high accuracy and meet the task analysis requirement of NAO robot by experiments.