US10665222B2 - Method and System of Temporal-Domain Feature Extraction for Automatic Speech Recognition