Autism Spectrum Disorder (ASD) is the most prevalent child neurological and developmental disorder causing cognitive and behavioral impairments. The early diagnosis is an urgent need for the treatment and rehabilitation of ASD patients. This work presents electroencephalogram (EEG) based ASD classification processor implementation that targets a patch-form factor sensor for long time monitoring in a wearable device. A patient is classified as ASD or typically developing using scalp EEG. The selection of frontal and parietal lobe electrodes causes minimum uneasiness to the children. The proposed and implemented algorithm utilizes only four EEG electrodes. The processor is implemented and validated on Artix-7 FPGA, requiring only 26229 lookup tables and 15180 flip flops. The hardware efficient implementation of the complex kurtosis value and Katz fractal dimension features using kurtosis value indicator and Katz fractal dimension indicator with 54% and 38% efficient implementations, respectively, is provided. A hardware feasible shallow neural network architecture is used for the ASD classification. The implemented system classifies the ASD with a high classification accuracy of 85.5% using the power and latency of 8.62 µW and 2.25 milliseconds, respectively.feeling of disgrace and repeated visits to neurologists. Fig. 1 shows the difference between the conventional ADOS-2 diagnosis and the proposed solution to solve the pain problem. The proposed solution (longer-term) would be able to diagnose a child as ASD or TD earlier in the form of a wearable head-band system on chip (SoC) processor. The processor would pre-process the brain waves (EEG), extract suitable features, and classify the child as ASD or TD using suitable machine learning (ML) or deep neural network (DNN) classification method with limited electrodes. It would not only avoid the stigma associated with the prolonged diagnosis but also reduce the rehabilitation costs due to early intervention [4].
