Matlab source codes for training and using the Integrated C-GMR (IC-GMR) model and the J-GMR model proposed in:

Hueber, T., Girin, L., Alameda-Pineda, X., Bailly, G. (2015) "Speaker-Adaptive Acoustic-Articulatory Inversion using Cascaded Gaussian Mixture Regression", in IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 23, no. 12, pp. 2246-2259

and

Girin, L, Hueber, T., Alameda-Pineda, X.,(2017) "Extending the Cascaded Gaussian Mixture Regression Framework for Cross-Speaker Acoustic-Articulatory Mapping", in IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 25, no. 3, pp. 662-673

CONTENT:

• cgmmTrain: Training routine for IC-GMR model (given a training set of N {input (x), output (y)} observations and an adaptation set of N0 {input (z)} observations with typically N0 << N)
• cgmmMap: Mapping routine for converting input (z) to output (y) observations using IC-GMR (MSE criterion)
• jgmmTrain: Training routine for J-GMR model (given a training set of N {input (x), output (y)} observations and an adaptation set of N0 {input (z)} observations with typically N0 << N)
• jgmmMap: Mapping routine for converting input (z) to output (y) observations using J-GMR (MSE criterion)
• gmmCalculateLikelihood & gmmCalculatePosteriors: subfunctions for calculating likelihood and posterior probabilities, adapted from GMR training routine of Dr. Sylvain Calinon (LASA Lab, EPFL).

CONTACT: Thomas Hueber, [email protected]