A5066542435- Music and Audio Processing
- Music Technology and Sound Studies
- Speech and Audio Processing
- Hearing Loss and Rehabilitation
- Multisensory perception and integration
National University of Singapore
2023
In this paper, we propose a data-driven approach to train Generative Adversarial Network (GAN) conditioned on "soft-labels" distilled from the penultimate layer of an audio classifier trained target set texture classes. We demonstrate that interpolation between such conditions or control vectors provide smooth morphing generated textures, and show similar better capability compared state-of-the-art methods. The proposed results in well-organized latent space generates novel outputs while...
Novel AI-generated audio samples are evaluated for descriptive qualities such as the smoothness of a morph using crowdsourced human listening tests. However, methods to design interfaces experiments and effectively articulate quality under test receive very little attention in evaluation metrics literature. In this paper, we explore use visual metaphors image-schema evaluate audio. Furthermore, highlight importance framing contextualizing measurement constructs. Using both pitched sounds...
In this paper, we propose a data-driven approach to train Generative Adversarial Network (GAN) conditioned on "soft-labels" distilled from the penultimate layer of an audio classifier trained target set texture classes. We demonstrate that interpolation between such conditions or control vectors provides smooth morphing generated textures, and shows similar better capability compared state-of-the-art methods. The proposed results in well-organized latent space generates novel outputs while...