An informed separation algorithm based on sound field mapping for speech recognition systems.