Designing a BCI Platform with Embedded ANN as an Aid for Autism Spectrum Disorder (ASD) Diagnosis: A Preliminary Study

This paper presents a technical approach to the design of a cost-effective BCI platform from the hardware construction based on commercial components to the firmware. This device may include drivers for signal acquisition, preprocessing, artifact removal, band separation, feature extraction, and classification via compressed ANN with the TensorFlowLite framework. It consists of an STM32-based EEG platform built-oriented as a tool for BCI experiments in the constraints of education and research fields, capable of sending data to PC or mobile devices via a wireless connection, allowing external processing capabilities alongside the main feature of embedded signal classification (TinyML) focused on early detection by discriminating children with ASD from non-ASD ones.