Research And Application Of Audio Recognition In Equipment Surveillance System

Industrial equipment will inevitably fail during operation, and it needs to detect and recognize the running state of the system equipments through surveillance system. Traditional surveillance system mainly uses field bus and video and image surveillance methods, but they are restricted in certain applications. Audio surveillance can make up for the deficiencies of video surveillance, and conducting the researches on the online fault detection of equipment audio signal has a practical value.Based on the profound study on the related theory and technology of audio recognition, thesis has designed a device state audio recognition system based on DSP TMS320DM643. Taking the measurement and control system of low-temperature plasma equipments as an application object, we have realized the function of real-time online survelliance on the device state by using audio recognition method to analyze equipment audio signal. The main contents are as follows:Firstly. Application background of device state survelliance system and related theory and technology of audio recognition were analyzed.Secondly, device state audio recognition system based on DSP TMS320DM643was designed. In view of the characteristics of low temperature plasma equipments’ running state, we extracted the audio siganl feature of vacuum pump, transformer;, high voltage power supply and some other different devices under different working condition. Thereafter, continuous density hidden Markov model was established. We designed the simulation program and established corresponding continuous density hidden Markov model. In the end, algorithm was transplanted into the DSP hardware platform, and we have successfully achieved real-time signal acquisition and processing, and have realized online audio recognition of low-temperature plasma device measurement and control system.Lastly. Audio data acquisition driver was created by the the software resources provided by TMS320DM643. Meanwhile, audio recognition algorithm was encapsulated according to XDAIS(eXpressDSP Algorithm standard), and RF5(Reference Framework5) algorithm framework for audio recogniton algorithm was set up. By using the real-time analysis tools provided by the development environment, we compared and analyzed the execution efficiency of the two algorithm proprams which adopted DSP/BIOS and RF5framework and the unadopted.