Andrés Ovidio Restrepo Rodríguez

In recent years, it has been shown that emotions influence what we learn and retain, and second-language learning is no exception to this phenomenon. Currently, a variety of mobile learning applications offer content for language learners, and a wide range of languages are presented. The analysis of emotional data in learning environments has been implemented through various methods, such as the collection of vital signs. This is where brain–computer interfaces (BCIs) play an important role in capturing emotional metrics from brain activity. Accordingly, this paper presents the Emotional Data L2 Interaction (EDaLI) dataset for emotional analysis based on the collection of emotions, such as engagement, stress, excitement, interest, relaxation, and focus, through Emotiv Insight, during the interaction of 19 participants with 4 initial lessons in Portuguese as a second-language, through the Babbel application. A preliminary visualization approach is proposed from the generated dataset. In accordance with this, it is concluded that visualization techniques can clearly be applied to EDaLI to show the emotional behavior exhibited by the participants during their interactions. Additionally, the spectrum of algorithms to be applied is open and includes possibilities such as the use of clustering techniques for time series of variable lengths.

Satellite image fusion provides a context for potential applications in several fields such as agricultural development, hydrology, environmental studies, and natural disaster actions plans. However, when dealing with large-size images, the time required for the fusion process grows significantly. To reduce processing delays, the present study proposes the use of the Brovey transform as an image-fusion technique together with a spectral richness calibration stage. The proposal makes use of a CPU/GPU heterogeneous computing architecture based on mass parallel processing, conducted with CUDA. The fusion process of a 8192-pixel image evinced a speed-up of 532X. Regarding the quality of the resulting image, a per-band average correlation coefficient of 0.9714 (spatial detail) was obtained when comparing the fused and panchromatic images in an (R,G,B) color space.

Fusion of satellite images consists of improving the quality of a multispectral image by combining data from a high spatial resolution panchromatic image with a high spectral resolution multispectral image. To carry out this, different techniques are available, which perform various operations at the pixel level, which leads to generating a dependency between the computational requirement and the image size. Currently, there are some libraries that implement these fusion methods, however, none of them allow this fusion process to be carried out on heterogeneous architectures, which enable the integration of acceleration platforms that reduce execution time through massive parallelization. For this reason, this document presents a library called Sallfus, which allows executing and evaluating the quality and performance of image fusion methods such as the Brovey transform, Multiplicative method, Principal Component Analysis (PCA) and Wavelet A trous, on homogeneous and heterogeneous architectures. Likewise, an evaluation of the library is made from an analysis of execution times and image quality using mathematical-statistical indices such as the correlation coefficient (CC), BIAS coefficient and Root of the Root Mean Square Error (RMSE). The results of the library evaluation showed that the merging process with images of 8192 pixels, presents a speed-up of approximately 591x for Brovey, 309x for Multiplicative, 18x for PCA and 6x for A trous. Additionally, it was observed that the methods that presented the best performance both computationally and in the quality of the merged image were Brovey and Wavelet A trous. Availability and implementation: https://github.com/Parall-UD/sallfus.

Immersive techniques such as augmented reality through devices such as the AR-Sandbox and deep learning through convolutional neural networks (CNN) provide an environment that is potentially applicable for motor rehabilitation and early education. However, given the orientation towards the creation of topographic models and the form of representation of the AR-Sandbox, the classification of images is complicated by the amount of noise that is generated in each capture. For this reason, this research has the purpose of establishing a model of a CNN for the classification of geometric figures by optimizing hyperparameters using Random Search, evaluating the impact of the implementation of a previous phase of color–space segmentation to a set of tests captured from the AR-Sandbox, and evaluating this type of segmentation using similarity indexes such as Jaccard and Sørensen–Dice. The aim of the proposed scheme is to improve the identification and extraction of characteristics of the geometric figures. Using the proposed method, an average decrease of 39.45% to a function of loss and an increase of 14.83% on average in the percentage of correct answers is presented, concluding that the selected CNN model increased its performance by applying color–space segmentation in a phase that was prior to the prediction, given the nature of multiple pigmentation of the AR-Sandbox.