Analysis on Food Crispness Based on Time and Frequency Domain Features of Acoustic Signal

Crispness is an important indicator of crunchy food. However, it cannot be easily quantified by sensory evaluation, due to the high subjectivity evaluators; instrument measurement this requires much manpower and time. To improve efficiency food crispness prediction, paper attempts build a rapid, convenient, accurate analysis model. Starting with fracturing sound food, authors collected acoustic signal, conducted wavelet denoising, analyzed eigenvalues in time frequency domains, constructed prediction models based on multiple linear regression (MLR) neural network, respectively. Through test test, cluster was adopted select typical including peak amplitude power spectral density (PSD) curve, difference, waveform index. Based these eigenvalues, model established, used predict four kinds namely, potato, sweet carrot, turnip. The results show that BP network had smaller relative error than MLR; when threshold 5%, maintained accuracy >90%, achieved 100% two types sum up, reveals relationship between chewing features quality, laying theoretical basis for research mechanism.