Investigation of self-supervised pre-trained models for classification of voice quality from speech and neck surface accelerometer signals