Whispered Speech

Whispered speech, distinct from normal speech due to the absence of vocal fold vibration, presents unique challenges for automatic speech recognition and human-computer interaction. Current research focuses on developing robust classification systems to distinguish whispered from normal speech, often employing convolutional neural networks (CNNs) and leveraging spectral features like the quartered spectral envelope to achieve high accuracy even in noisy environments. These advancements, along with innovative microphone designs and generative models for speech conversion and enhancement, aim to improve the reliability and usability of whispered speech in applications ranging from privacy-preserving voice interactions to assistive technologies. The ultimate goal is to create systems that seamlessly integrate whispered speech into everyday technology.