Here we prove the theoretical limit of a method implemented in my paper: Voice Response-based Emotion Intensity Classification for Assistive Robots
An important asssumption we made here is that each audio segment were independent and identically distributed (i.i.d), meaning that,
1) All four emotion intensity each follow a distinct probability distribution and an audio segment belongs to only one of any probability distribution (Identical Criteria)
2) Each audio segment does not influence the other (Independent Criteria)