Lionel

Builiding an "autocorrect for speech pathways" device using machine learning and transcranial direct current stimulation - With help from friends at Imperial College London I am building a machine learning triggered autocorrect for speech device using transcranial direct current stimulation and neural networks to increase fluency in people who stutter. The device is able to detect when the stutter occurs and immediately deliver a dose of stimulation to certain part of the brain. - Supported by several studies my thesis is that using the device in a live environment will rewire the brains speech pathways and overtime increase verbal fluency. This device may translate to other conditions such a parkinsons, or potentially slowing age induced cognitive decline. - Finally I am using the latest advancements in generative AI to build the first of its kind synthetic audio data generator which replicates the speaking patterns of those who stutter and later other speech impediments. - The lack of quality data in the field is one of the largest bottlenecks, mentioned in almost every paper. By being able to generate a near unlimited amount of such data, whole new frontiers may open up. The project is still very much research based but soon a physical prototype will be available. This idea was conceived 5 years ago and the field of AI and ML has finally caught-up.