Ez egy előző félévben kiírt, archivált téma.
Speech sounds are produced as the coordinated movement of the articulatory organs (vocal folds, tongue, lips, etc.). In the international research, the machine learning based articulatory-to-acoustic mapping is also referred as 'Silent Speech Interface' (SSI, https://t.co/M3yY4WnMmF). The main idea of SSI is that by recording the speechless movement of the speaking organs, the algorithms can synthesize speech while the original speaker is not producing any speech sound. The task of the student is to develop deep learning (e.g. deep neural network, AutoEncoder) solutions for the Silent Speech Interface topic. Suggested programming languages: Python / Octave. The research will be conducted in collaboration with the MTA-ELTE Lingual Articulation Research Group (Momemtum grant, http://lingart.elte.hu/en).