Sound Source Separation And Localization Of High Pressure Water-jet Target Reflective Sound Signals

Detecting mines is full of danger and challenging. It is experimentally verified that the high pressure water-jet cutting the mines is efficient and safe. Therefore, the mines detection and mines exclusion can be integrated by combining high pressure water-jet technology, microphone array and sound identification, which improves the efficience and security of the mines detection. The mines detection essence is targets recognition. To realize the function, the first step is separating the environment noise from the target refelective sound signals produced when the targets are attacked by high pressure water-jet. On the other hand, it is a necessary to locate the sound source accurately and in real time after the targets identification.The sound source separation and soud source localization related to the mines detection have been done the following jobs in the paper. The present research states at home and abroad of high pressure water-jet technology, sound source separation and sound source localization were analyzed. The theory of independent component analysis and sound source localization based on time difference was described. The hardware system was composed of microphone array, data acquisition card, computer interface and so on. The measurement software of independent component analysis and sound source localization was programmed. The localization formulas based on linear microphone array were deduced. The simulation experiments about FastICA belonged to independent component analysis and cross-power spectrum phase time delay estimation of sound source localization based on time difference were done. The indoor and outdoor experiment datas were acquired. What’s more, the two experiment datas were both processed with the sound source separation and sound source localization.The simulation and experiment results show that the environment noise can be separated from the different targets reflective sound signals effectively by FastICA, which reconstructs the detecting-line sound distribution, and that the time delay between the microphone array can be estimated approximately by cross-power spectrum phase time delay estimation. The sound source localization was realized by the deduced localization formulas too. All of the results provide premises to the progress of the project and do spade work for the eigenvalues extraction of reflective sound signals and target recognition.Figure [42] table [8] reference [58]...