Research On The Sound-Fault Detection Method For Micro Loudspeaker

Fault detection of loudspeaker is the key factor which deciding to produce quality in the production process. The main way to detect loudspeaker is human ear monitoring currently, but some subjective and objective factors such as sentiment of workers, working ambient, have an effect on fault detection. With the development of information technology, loudspeaker has a wider range of applications, which makes the development of detection method tend to be digital and intelligent. At present, the studies of loudspeaker have already had a certain depth at home and abroad, but there are a few studies about micro loudspeaker. The characteristics of micro loudspeaker, for example low power, small volume, which make a different with normal loudspeaker in failure performance. So according to micro loudspeaker’s own feature, the fault detection research has become electro-acoustic industry’s urgent needs on micro loudspeaker.At present, the detection systems are simultaneously applied to micro loudspeakers and common loudspeakers. Then audio signal generator triggers micro loudspeakers, microphone gets sound-pressure signal, through A/D converter, data collection card, computer gets response signal. Finally, it realizes fault identification by JTFA (Joint Time-Frequency Analysis) and characteristic values (or characteristic curve). In view of micro loudspeaker own characteristics, the thesis proposes a way based on time-domain segmentation to detect fault micro loudspeaker. And according to physical characteristics and power of micro loudspeaker, it designs anechoic chamber for obtaining more obvious characteristics.The main task and research results including:1. According to the relation of time-frequency resolution, the thesis proposes a detection method based on time-domain segmentation.2. This thesis uses the method of short-time Fourier transform for the analysis of the loudspeaker’s response signal, and proposes an Image Euclidean distance method for the feature extraction.3. It designs an anechoic chamber, the cabinet and the trestle of microphone.4. The thesis completes the integration on hardware and software systems.The experiments show that the fault detection method based on time-domain segmentation could well detect the fault micro loudspeaker and distinguish bottoming fault, leaking fault.