Speech/music classification using phase-based and magnitude-based features