This invention relates to a means for compensating for a whole or partial loss of speech. The means has a sensor unit in the form of a Logometrix pseudo palate and a palatometer which sense the position of a user's tongue while endeavouring to articulate words or sounds. The sensor unit is connected to a comparator which makes use of suitable artificial intelligence algorithms (such as Multi-Layer Perceptron Neural Networks and Hidden Markov Models) together with a library of stored words and sounds as well as tongue positions and signals related to jaw and lip movement as well as inhalation and exhalation to match words or sounds and transmit the matched words or sounds to a loudspeaker. Preferably the output of the loudspeaker will simulate the natural voice of the user.
