Study And Design Of Speech Interactive System For Robots

With the development of multimedia technology, the key technology is communicating between human and robots. The voice is one of the most important, availability and convenient method to convey information from one to another. Combing the both which make voice transmission, storage information, and human-robots interaction by speech has become an important research topic. Adding the voice interaction function to computer, by which can make human-computer interaction naturally and friendly just like interpersonal interaction. This natural and harmonious interaction technique has become one of development trends in the field of human-computer interaction.We are designing the robot of BME-1which serves to humanity; we design the speech interactive system to achieve communicating between human and robots. With the communication process of human beings and robots through the language, we can achieve the requirements of the interactive and scientific. This thesis designs a system in detail and realizes the initial function of the system based on the research and analysis about function and key technique of the voice interaction system.Firstly, this thesis introduces the speech Interactive system which has three modules briefly, the speech analysis module, the speech recognition module and the speech synthesis module. By the way, the speech analysis module includes the speech enhancement processing and the feature parameter extraction of the speech signals.Secondly, though anglicizing and researching on the different modules, we know that the front-end processing of the speech signals include speech signal Pre-emphasis, framing and window treatment. And then calculate to get the appropriate microphone array, according to the system requirements, we select the circular array, and studied the algorithm of the speech enhancement. In this thesis, we selected the algorithm based on Time domain constrained (TDC) Estimator. In the analysis section of the characteristic parameters of the speech signal, we give the method of endpoint detection and pitch period extraction of the speech signals. We mainly use the dual threshold endpoint detection and the pitch extraction based on Short Time Autocorrelation Function (AMD), Short Time Average Difference Function (AMDF). And smooth the speech signals make sure the voiceless and the voiced separated by the second judgment.Thirdly, we design the speech interactive system based on Microsoft Speech SDK. The speech recognition is designed using speech application program interface provided by Speech SDK. Recognize result is dealt with in logical program. Thus, voice chat between human and robots can be realized.