Speech coding using phonetic or linguistical decoding of the source; Reconstruction using text-to-speech synthesis