Data-driven Approaches for Discovery and Prediction of User-preferred Picture Settings on Smart TVs

Abstract

We discover user-preferred picture settings on smart TVs and investigate whether it is possible to predict the users' picture setting preferences through machine learning methods. We frst perform K-means clustering on large-scale smart TV usage log data to understand how users fne-tune the factory default picture settings. Clustering results recognize 3-4 user groups who have reasonably diferent preferences toward the default settings. By characterizing these user preferences, we come up with new user-preferred picture settings. We perform an in-depth analysis of the newly discovered picture settings regarding diverse TV device characteristics. We also perform lab experiments to demonstrate how these new settings deliver diferent picture quality than the default. Next, we construct a deep learning-based classifer that learns and predicts the picture setting preferences of the users. The fnal trained model shows 86% accuracy in predicting users' decisions to choose a specifc picture setting out of four available options. CCS CONCEPTS• Information systems → Personalization; • Computing method ologies → Cluster analysis; Supervised learning by classifcation; • Applied computing → Consumer products.

Evaluation of Subjective Video Quality of Television Displays

“…The subjects tend to prefer the bright Movie 2; p < .005). This finding sounds reasonable based on the literature: people prefer brighter images on a display device (e.g., TV) in a bright environment and vice versa [16,17].…”

“…Toward this end, we designed two experiments to evaluate the impact of varied combinations of select factors on specific video quality, implemented by a TV display's picture setting configurations and an image processing algorithm. In the first experiment, we asked subjects' preferences regarding specific picture settings we developed based on TV usage log data [16,17]. These settings differ from the factory default, mainly on brightness, contrast, and sharpness lev-els.…”

“…The original video format was Apple's QuickTime (MOV), and we encoded it using HEVC for easy content playback on TVs. As discussed before, we prepared specific picture settings (also known as Picture Mode) derived from analyzing TV usage log data [16,17]. Here, we focused on two popular Picture Modes, Standard and Movie.…”