Short answers, pulled from the story.
Machine perception is the capability of a computer system to interpret data in a way similar to how humans use their senses. It differs from broader artificial intelligence in that machine perception aims to grant machines limited sentience rather than full consciousness, self-awareness, and intentionality.
The main types are computer vision, machine hearing, machine touch, and machine olfaction. Each field processes a different category of sensory input, from images and sound to tactile information and airborne chemicals.
Auditory scene analysis is the ability of a machine to selectively focus on a specific sound against many other competing sounds and background noise. The technology enables the machine to segment several audio streams occurring at the same time.
Scientists have not yet invented a mechanical substitute for nociceptors, the structures in the human body and brain responsible for noticing and measuring physical pain and discomfort. Current tactile research focuses on combining tactile sensors with machine learning to handle tasks like robotic surgery and prosthetics with sensory feedback.
Moravec's paradox is a named unsolved problem listed among the future hurdles machine perception research must overcome. It is one of several challenges, alongside embodied cognition and the principle of similarity, that researchers identify as barriers to machines achieving human-like sensory interpretation.
The end goal of machine perception is to give machines the ability to see, feel, and perceive the world as humans do. This would allow machines to explain their decisions in human terms, warn when they are failing, and state the reason for that failure.