Acoustic Performance Analysis of Micro Speaker Using Machine Learning Algorithms

摘要

Micro speakers are electro mechanical transducers that converts an electrical signal into sound. Micro speaker dimensions are in the range of20mm or even less for at least one dimension. Micro speakers are generally used in speech assist systems, cars, mobile devices, smart speakers. While the video screens of cell phones, notebooks and tablets have seen enhancements, there exists a lag in the improvement of audio performance. Micro speaker in cell phone is limited by small size, still sound tinny and quiet. Architects use different techniques to increase the sound quality and volume, but with limited achievement. The output of micro speakers is modeled in the echo cancelation process to improve the voice quality by preventing echoes from being captured or created, or possibly removing it in post-processing. However blown speakers are a typical cause of failures in cell phones and other devices. The use of laser technology to identify the defects in the micro speaker during their usage in the various products is unrealizable and are very expensive. Hence we propose an automated method to analyze the performance of micro speakers in the real time using objective speech qualitive assessment techniques and machine learning algorithms(ML).