Datasets for Automated Affect and Emotion Recognition from Cardiovascular Signals Using Artificial Intelligence- A...

No prior work had resolved the limitations regarding data transparency in physiological emotion recognition. That uncertainty drove this investigation into the current state of publicly accessible information. Prior research has shown that machine learning models rely heavily on high-quality input data. This gap motivated an analysis of existing repositories for heart-based affective computing. It was already known that diverse signals exist for monitoring human emotional states. However, the consistency of these resources remained largely unexamined by the scientific community. This study addresses how researchers manage cardiovascular data for automated systems. Such efforts are required to ensure that future technological developments remain robust and replicable.

The aim of this review was to assess the current state and quality of publicly available datasets for automated affect and emotion recognition. This investigation specifically emphasized the use of cardiovascular signals within artificial intelligence frameworks. The authors sought to determine if existing resources are sufficient to create replicable systems for future technological growth. A primary motivation was the observation that high-quality data is required for reliable model development. The researchers addressed the problem of inconsistent reporting standards across the scientific literature. By examining these datasets, the study highlights the challenges faced by developers in this domain. This work provides a critical overview of how physiological data is currently managed and shared. The authors intended to provide a foundation for improving future data collection and documentation practices.

Review approach involved a systematic search across nine distinct databases up to August 2020. Two independent reviewers performed the screening of records to ensure consistency throughout the selection process. The team conducted a full-text assessment and extracted data from the identified studies. All discrepancies were resolved through discussion to maintain high standards of accuracy. The protocol was registered on the Open Science Framework platform to promote transparency. Researchers descriptively synthesized the results to provide a clear overview of the current landscape. This design allowed for a comprehensive evaluation of the quality and credibility of existing resources. The methodology focused on identifying datasets that specifically utilized heart-based signals for emotion detection.

Key findings from the literature indicate that 18 records were selected from an initial pool of 4649 studies. These included papers analyzed data from 812 participants aged between 17 and 47 years. Electrocardiography served as the most explored signal type, representing 83.33% of the identified datasets. Regarding stimulation techniques, video content was utilized most frequently in 52.38% of the experiments. Despite these figures, the analysis revealed that much information was not reported by the original authors. The overall quality of the examined papers was determined to be mainly low. These results demonstrate a significant gap in the documentation of experimental procedures. Consequently, the findings underscore the urgent need for improved methodological standards within the field.

The authors suggest that the overall quality of existing datasets remains suboptimal for reliable model development. Synthesis and implications indicate that researchers must prioritize rigorous methodological reporting in future publications. The review highlights that many studies fail to provide sufficient information for independent verification. This lack of transparency hinders the progress of automated systems in this domain. The researchers propose that standardized protocols could improve the utility of shared physiological data. Future work should focus on enhancing the documentation of participant demographics and experimental conditions. These findings imply that the field currently lacks a unified approach to data curation. Improving these practices will be necessary for the long-term advancement of affective computing technologies.