Speech Analysis for Autism Spectrum Disorder Detection for Children

People with Autism Spectrum Disorder (ASD) will have problems with social interaction and communication. This may impair their quality of life. Currently there is no medicine for treating ASD. Fortunately, its symptoms may be alleviated if early interventions are applied. Early screening for ASD is therefore essential. Proposed methods in literature used behavioral and physiological signals to deal with this problem. In our research, we used four features extracted from speech signals of three- to six-year children. These were tested with seven classification models. Experiments showed that these features of speech, combined with decision tree, can be employed for early screening of ASD for children with an error rate of 0.25%.