This paper delves into the complexities of autism spectrum disorder (ASD) diagnosis and therapeutic interventions, emphasizing the critical role of occupational therapy (OT) in addressing the diverse challenges faced by individuals with ASD. Depicting the diagnostics, therapy, and the motivation behind identifying social cues are pertinent to paint a picture of the various aspects involved in managing autism spectrum disorder (ASD). Also, a systematic literature review evaluates the applicability of deep learning methods in connection with different behavioral features such as facial motions, gaze information, motor signs, and repetition, in the context of ASD. The paper also offers a distinctive deep learning approach intended for examining video sequences included in the analysis of children with ASD and their ability to recognize emotions. Last but not the least, the results and discussion section focuses on positive results of the introduced model, and further, the limits of this model for the intervention of ASD explain the importance of a better understanding of the emotion recognition for the improvement of the therapeutic processes.