 This paper proposes a novel multimodal diagnosis framework for Autism Spectrum Disorder, ASD, in children. The authors design two stacked denoising autoencoder, SDAEs, models for feature learning for electroencephalography, e.g., anti-tracking, ET, modalities, respectively. A third SDAE model is then used to fuse these two modalities together in order to improve the accuracy of diagnosing ASD. This approach was tested on a dataset of 40 ASD children and 50 typically developing, TD, children, and the results show that the proposed method outperformed other unimodal methods and a simple feature-level fusion method. This article was authored by Jun Xiaohan, Guqin Jiang, Gao Xiang Ouyang, and others.