Use of Multi-Layered Networks for Coding Speech with Phonetic Features