Ultraspeech-acquisition is a software application that allows the simultaneous and synchronous acquisition of ultrasound images (up to 70 fps in 640x480 pixels), video images (up to 60 fps in 640x480 pixels NB), multiple audio channels and inertial data from accelerometers. Ultraspeech-acquisition is compatible with Terason T3000 and Telemed ultra-portable ultrasound scanners (Echoblaster and MicroUS), industrial cameras from Imaging Source, and ASIO compatible sound cards. Free download at !


Check out these sequences recorded using Ultraspeech !




Intuitive visualization of ultrasound articulatory data for speech therapy and pronunciation training


Ultraspeech-player is a standalone software dedicated to the visualization of ultrasound speech data recorded using Ultraspeech. Ultraspeech-player is designed for pronunciation training in the context of speech therapy and second language learning. The software aims at displaying natural tongue movements acquired on a reference speaker, for different kind of sequences (isolated vowels, VCV, swallowing, etc.). It also includes an audiovisual time-stretching module allowing the user to slow-down both the articulatory gesture and its corresponding acoustic realization (i.e. the speed of the audio signal is modified while the original pitch it preserved). This rendering technique aims at improving the way a naïve speaker perceive and understand a tongue gesture. Free to download from Ultraspeech-player webpage !

Codes sources / toolbox

Cascaded-GMR: Matlab toolbox for training and using the C-GMR models (SC-GMR, IC-GMR and J-GMR) proposed in (Hueber et al., IEEE/ACM TASLP 2015) and (Girin et al., 2016 IEEE/ACM TASLP)

DeepPredSpeech:  Python script for training and using the computational model of speech predictive coding proposed in  (Hueber et al., 2019, Neural Computation).



Grenoble Images Parole Signal Automatique laboratoire

UMR 5216 CNRS - Grenoble INP - Université Joseph Fourier - Université Stendhal