ASD is a Neuro development disorder. Lack of communication, anxiety over social interaction, feeble and confused behaviors are some of the outward impacts and intense impressions found in an autistic patient. Early detection and diagnosis of autism can feasibly eliminate the progression of the disease. In this paper, the detection of autism is elevated by incorporating facial expression analysis along with Machine learning techniques. The goal of the system is to evaluate the facial expressions of the patient by interpreting images or videos which aids to figure out the potential indicators of ASD. Data collection embroiled with diverse dataset that encompasses facial expression of the neurotypical individuals. In order to enhance the data quality few valid preprocessing techniques are incorporated a prompt facial expressions, facial features and expression patterns are implied as feature extraction methods. Multi-dimensional spatial-temporal sparse based convolutional long short-term memory is engaged to incriminate the neuro typical behavior of individuals and to analyze their facial expressions. The proposed model gets trained by the preprocessed dataset. The dataset is inculpated by proper validation and testing which leads to examine its real-world performance.