Avatar Audio-Visual Evaluation Using Automated Naturalness Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for evaluating virtual avatars rely heavily on human intervention, leading to inefficiencies, errors, and high costs due to the need for skilled labor, and do not effectively ensure the quality and naturalness of virtual avatars in interactive environments.
Innovation Solution
An automated evaluation system that combines audio and video evaluation metrics to assess the naturalness and quality of avatars, using an evaluation module with audio and video evaluators to generate a combined naturalness score for the avatar generator, eliminating the need for human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual evaluation systems are used to assess virtual avatars, then evaluation accuracy can be maintained through human judgment, but the process becomes time-extensive and costly due to required skilled labor
Solution Approach 1:
The patent replaces the mechanical human evaluation system with an automated computational system that uses audio and video evaluation metrics to assess avatar naturalness. The system substitutes human listeners and visual assessors with algorithms that automatically compare avatar audio-visual outputs against reference data, eliminating the time-consuming manual process while maintaining evaluation capability through programmed assessment criteria
Solution Approach 2:
The patent creates computational models and reference profiles that copy the evaluation capabilities of human assessors. By capturing and replicating human judgment criteria in software form, the system can perform repeated evaluations without the time loss associated with recruiting and training new human evaluators, while preserving the precision of human-like assessment through carefully designed metric comparisons
2Reliability
If manual evaluation systems are used to assess virtual avatars, then nuanced judgment can be applied, but the cost increases due to required skilled labor
Solution Approach 1:
The patent replaces expensive human expert systems with automated computational evaluation that uses audio and video metrics. This substitution eliminates the need to recruit, train, and compensate skilled human evaluators, dramatically reducing costs while maintaining reliable assessment through systematic comparison of avatar outputs against established reference profiles and evaluation criteria
Solution Approach 2:
The evaluation system is designed to be self-sufficient, automatically performing assessments without requiring human intervention for each evaluation. The system self-calibrates using reference data, automatically compares avatar outputs against multiple metrics, and generates evaluation results independently, eliminating the ongoing cost of skilled labor while maintaining consistent evaluation quality
3Productivity
If automated evaluation systems are implemented, then efficiency and cost-effectiveness improve, but the complexity of the system increases
Solution Approach 1:
The patent divides the evaluation system into distinct modular components: an audio evaluation module that assesses speech characteristics, a video evaluation module that analyzes visual avatar output, and a naturalness scoring module that integrates results. This segmentation allows each component to be independently developed and optimized, managing overall system complexity while enabling high-throughput parallel processing that improves evaluation efficiency
Solution Approach 2:
The evaluation system is designed as a universal platform that can assess multiple types of avatars using the same audio and video evaluation metrics. By creating a multi-functional system that handles various avatar formats and evaluation scenarios through common processing pipelines, the patent achieves high productivity across diverse applications without proportionally increasing system complexity
4Measurement precision
If extensive human intervention is used in avatar evaluation, then evaluation thoroughness is maintained, but the process becomes time-extensive
Solution Approach 1:
The patent replaces sequential human evaluation processes with parallel automated computing operations. Multiple audio and video metrics are simultaneously calculated and compared against reference profiles, maintaining thorough assessment of avatar naturalness while achieving high evaluation throughput through concurrent processing of multiple measurement dimensions that would be sequentially performed by human evaluators
Data Source
AI summary
A system obtains, by an audio evaluator, a speech generated by a text-to-speech module of the avatar generator. The system obtains, by the audio evaluator, audio features of a target person that the avatar is representing. The system compares the speech with the audio features of the target person using a set of audio metrics, and generating an audio evaluation score for the speech based on a comparison of the speech and the audio features, wherein generating the audio evaluation score comprises evaluating one or more of: m speech intelligibility using automatic-speech-recognition (ASR) based evaluation metrics, audio noise level using voice-activity-detection (VAD) based evaluation metrics, naturalness of speech intonation using pitch-based metrics, voice similarities using equal-error-rate (EER) and cosine (COS) metrics, and speech pronunciation statistics.