Affect-Based Video Recommendation via Mental State Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video recommendation systems rely on imprecise and subjective star ratings, which are tedious for users and fail to accurately capture individual and collective responses to videos, leading to unreliable recommendations.
Innovation Solution
A computer-implemented method that captures mental state data, including physiological and facial data, while playing a media presentation, and recommends subsequent media based on this data by comparing it to mental state event temporal signatures and aggregating data from other viewers, providing affect-based recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If star ratings are used for video recommendations, then users can provide feedback, but the ratings are imprecise and subjective leading to unreliable recommendations
Solution Approach 1:
The patent replaces the mechanical/manual star rating system with an automated affect analysis system that uses facial expression recognition, physiological sensors, and machine learning algorithms to objectively measure viewer emotional states, thereby substituting subjective mechanical feedback with objective automated measurement
Solution Approach 2:
The patent changes the measurement parameters from discrete star ratings to continuous affective dimensions including valence, arousal, dominance, and specific emotions, allowing for much finer-grained and more precise measurement of viewer responses to video content
2Ease of operation
If star ratings are used, then users can evaluate videos, but the process is tedious for users
Solution Approach 1:
The system allows viewers to passively provide feedback through automatic affect analysis of their facial expressions and physiological responses without requiring active participation, making the evaluation process effortless while maintaining high measurement precision through automated emotion recognition
Solution Approach 2:
The patent replaces the manual mechanical act of selecting star ratings with automated biosensing and facial analysis systems that continuously and passively measure affective states, eliminating the need for user intervention while capturing precise emotional responses
3Reliability
If traditional recommendation systems are used, then recommendations can be generated, but they fail to accurately capture individual and collective responses to videos
Solution Approach 1:
The patent transforms the information parameters from simple rating values to comprehensive affective profiles including multiple emotional dimensions (valence, arousal, dominance), specific emotions (joy, sadness, anger, fear), and temporal patterns, thereby capturing rich emotional response information that traditional systems lose
Solution Approach 2:
The patent adds new dimensions to the recommendation space by incorporating temporal evolution of affective states and multi-dimensional emotional profiles, moving beyond single-point ratings to capture the dynamic and nuanced nature of human emotional responses to video content
Data Source
AI summary
Analysis of mental state data is provided to enable video recommendations via affect. Analysis and recommendation is made for socially shared live-stream video. Video response is evaluated based on viewing and sampling various videos. Data is captured for viewers of a video, where the data includes facial information and/or physiological data. Facial and physiological information is gathered for a group of viewers. In some embodiments, demographic information is collected and used as a criterion for visualization of affect responses to videos. In some embodiments, data captured from an individual viewer or group of viewers is used to rank videos.