Non Intrusive Speech

Non-intrusive speech quality assessment (NISQA) aims to objectively evaluate speech quality without needing a clean reference signal, a significant advantage over traditional methods. Current research focuses on developing and improving deep learning models, particularly convolutional neural networks (CNNs), recurrent neural networks (RNNs), and transformer-based architectures, often incorporating techniques like multi-task learning and self-supervised learning to enhance prediction accuracy and robustness. This field is crucial for applications like online conferencing and hearing aid technology, enabling real-time monitoring and optimization of speech communication systems and improving the assessment of speech intelligibility for individuals with hearing impairments.

Papers