Fig. 3
From: Multimodal dataset of real-time 2D and static 3D MRI of healthy French speakers
Data processing routine illustrated with the data of P9. The figure should be read from left to right, and each item corresponds to a processing stage. The first item lists the acquired data, second and third items illustrate sound alignment and denoising results, and the last item aims to explain the manual alignment principles. The articulators’ position on the MR image 1265 corresponds to the onset of /p/ (the lips have just come in contact). The plot below the MR image shows a spectrogram of the sound with vertical lines denoting centres of MRI acquisition intervals (each 5-th line is black for better readability). From the spectrogram, the image 1265 is supposed to correspond to the middle of /p/, which is not the case. The sound is haste by approximately 3 images = 60 ms and thus the sound and the images should be re-aligned.
