Taking advantage of machine learning and pattern recognition in acoustic measurements

摘要

Traditionally, acoustics measurement researchers and developers have focused on: a) the development of novel instruments and sensors, trying to improve their accuracy or trying to fix specific problems deriving from specific applications; b) the definition of uncertainty estimation methodologies, either in instruments and testing methods, c) the definition and customization of new measurement indexes. In very recent years, the research lines have been changing, in an attempt to take advantage of new signal processing techniques and solve classical handicaps in acoustic measurements. This is the case, for example, of blind source separation or beamforming. Following this trend, this paper focuses on new acoustic measurement approaches deriving from the application of machine learning and pattern recognition techniques for the development of smart instruments based on the measurement of sound pressure level.