Sensing technologies for silent speech interfaces

Silent speech interfaces decode speech intent without audible sound, enabling communication in settings where voice is inaccessible, or for individuals with speech impairments. Here we examine how sensing technologies shape the capabilities of silent speech interfaces. We compare off-, on- and in-body sensing modalities, identifying how proximity, coupling stability and invasiveness govern signal fidelity, robustness and user comfort. We highlight key trends, including the rise of flexible bioelectronics, multimodal sensor fusion for artefact resilience, and the growing role of edge artificial intelligence in real-time, low-power decoding. We show that on-body systems currently offer the best balance between accuracy and deployability, whereas in-body approaches provide unmatched neural access for individuals with complete loss of articulation. Looking ahead, advances in multimodal sensing, embedded intelligence and closed-loop architectures are poised to expand silent communication across rehabilitation, daily interaction and human–machine interfaces. Silent speech interfaces rely on a wide range of sensing technologies to decode articulatory and neural signals without sound. This Review compares off-, on- and in-body modalities and outlines emerging sensor–AI integration trends for scalable silent speech interfaces.