A Brain machine interface (BMI) is an emerging field of computer science and neurology that has shown expediency in various escalating and established applications across distinct scientific disciplines. Several neural monitoring techniques based on brain imaging and neural signals are utilized for different BMI applications such as mental states such as emotion and stress and mental disorder/diseases such as autism, cerebral palsy, myotrophic lateral sclerosis, stroke, or spinal cord injury. This article focuses on a comprehensive review of recent BMI techniques for emotion recognition and stress detection. It provides extensive description of various neural signals, methodologies for learning of neural signals, advantages, and disadvantages. Based on the comprehensive survey this article proposed potential challenges and constraints of neural monitoring techniques that needed to be considered in future.