Life Period Estimation of Stamping Process Using Punch Sounds and Deep Neural Network

This paper presents the life period estimation of stamping process by analyzing its punch sounds with a deep neural network (DNN). Neuro-approximated MFCC (NA-MFCC) feature extraction has also been proposed to trade estimation accuracy for computation complexity at a finer granularity. Several Pareto-optimal configurations have been reported with the performance measured on an ARM Cortex A15-based prototype.