Intelligent voice quality assessment posttreatment using genetic programming

Objective techniques for assessing and classifying voice quality for patients recovering from treatment for cancer of the larynx have largely focussed on they use of Artificial Neural Networks (ANN). The results of a preliminary study are reported, where a Genetic Programming (GP) has been trained to classify recovered (normal) and abnormal voices in acoustic data, and produced much more accurate results than an ANN. In addition, the GP is able to provide impact factors for the various voice parameters, and suggests that only 6 of the 22 short-term and longterm parameters used in the current ANN studies are contributing significantly to the classifications.