Robust Deep Learning Frameworks For Acoustic Scene And Respiratory Sound Classification