Multi-modal expressive personality recognition in data non-ideal audiovisual based on multi-scale feature enhancement and modal augment

Personality has long been the of psychologists' researchfocus, and it is a comprehensive expression of individual psychological traits that permeate human cognition, emotion and behaviour. From a psychological perspective, personality not only reflects a person's specific response patterns in a stable environment, but also embodies the complex interweaving of his or her intrinsic motivations, values and worldview. From another perspective, the influence of personality is ubiquitous in human society, cutting across a wide range of domains, including interpersonal relationships, career choices, educational strategies, and mental health. For example, people who are more extroverted are more inclined to engage in occupations that require frequent social interactions, while people who are more dutiful usually show higher levels of self-discipline and task completion. In addition to this, first impressions of an unfamiliar face can be generated in less than 100 milliseconds of contact based on personality traits[1], and first impressions play an important role in everyday life, such as interviews, elections, and dating. Various approaches and theories have been developed to classify, explain and measure personality. Vinciarelli et al. argue that personality is a psychological construct whose main purpose is to explain the diversity of human behaviour and which [2]can be predicted by stable and measurable characteristics of individual human beings. Costa et al. argue that traits are the key elements in the make-up of personality, and are the used to [3]basic units of personalitymeasurement measure and assess, and argue that human habitual patterns of and emotions are relatively stable over time.