Enhancing Sound Texture in CNN-Based Acoustic Scene Classification