Randy F. Fela

Perceptual evaluation of immersive audiovisual quality is often very labor-intensive and costly because numerous factors and factor levels are included in the experimental design. Therefore, the present study aims to reduce the required experimental effort by investigating the effectiveness of optimal experimental design (OED) compared to classical full factorial design (FFD) in the study using compressed omnidirectional video and ambisonic audio as examples. An FFD experiment was conducted and the results were used to simulate 12 OEDs consisting of D-optimal and I-optimal designs varying with replication and additional data points. The fraction of design space plot and the effect test based on the ordinary least-squares model were evaluated, and four OEDs were selected for a series of laboratory experiments. After demonstrating an insignificant difference between the simulation and experimental data, this study also showed that the differences in model performance between the experimental OEDs and FFD were insignificant, except for some interacting factors in the effect test. Finally, the performance of the I-optimal design with replicated points was shown to outperform that of the other designs. The results presented in this study open new possibilities for assessing perceptual quality in a much more efficient way.

For accurate and detailed perceptual evaluation of compressed omnidirectional multimedia content, it is imperative for assessor panels to be qualified to obtain consistent and high-quality data. This work extends existing procedures for assessor selection in terms of scope (360? videos with high-order ambisonic), time efficiency, and analytical approach, as described in detail. The main selection procedures consisted of a basic audiovisual screening and three successive discrimination experiments for audio (listening), video (viewing), and audiovisual using a triangle test. Additionally, four factors influencing quality of experience, including the simulator sickness questionnaire, were evaluated and are discussed. After the selection process, a confirmatory study was conducted using three experiments (audio, video, and audiovisual) and based on a rating scale methodology to compare performance between rejected and selected assessors. The studies showed that (i) perceptual discriminations are influenced by the samples, the encoding parameters, and some quality of experience factors; (ii) the probability of symptom occurrence is considerably low, indicating that the proposed procedure is feasible; and (iii) the selected assessors performed better in discrimination than the rejected assessors, indicating the effectiveness of the proposed procedure.

In an earlier study, we gathered perceptual evaluations of the audio, video, and audiovisual quality for 360 audiovisual content. This paper investigates perceived audiovisual quality prediction based on objective quality metrics and subjective scores of 360 video and spatial audio content. Thirteen objective video quality metrics and three objective audio quality metrics were evaluated for five stimuli for each coding parameter. Four regression-based machine learning models were trained and tested here, i.e., multiple linear regression, decision tree, random forest, and support vector machine. Each model was constructed using a combination of audio and video quality metrics and two cross-validation methods (k-Fold and Leave-One-Out) were investigated and produced 312 predictive models. The results indicate that the model based on the evaluation of VMAF and AMBIQUAL is better than other combinations of audio-video quality metric. In this study, support vector machine provides higher performance using k-Fold (PCC = 0.909, SROCC = 0.914, and RMSE = 0.416). These results can provide insights for the design of multimedia quality metrics and the development of predictive models for audiovisual omnidirectional media.