Application of E-aNets to Feature Recognition of Manner of Articulation in Knowledge-based Automatic Speech Recognition