APPROACHES TO ENVIRONMENT COMPENSATION IN AUTOMATIC SPEECH RECOGNITION