Improving Self-Supervised Learning for Audio Representations by Feature Diversity and Decorrelation | IEEE Conference Publication | IEEE Xplore