MIXTURE DENSITY NETWORKS, HUMAN ARTICULATORY DATA AND ACOUSTIC-TO-ARTICULATORY INVERSION OF CONTINUOUS SPEECH
Korin Richmond · 2024
Researchers have been investigating methods for retrieving the articulation underlying an acoustic speech signal for more than three decades. A successful method would find many applications, for example: low bit-rate speech coding, helping individuals with speech and hearing disorders by providing visual feedback during speech training, and the possibility of improved automatic speech recognition.