The classification of surface myoelectric signals (sEMG) remains a great challenge when focused on its implementation in an electromechanical hand prosthesis, due to its nonlinear and stochastic nature, as well as the great difference between models applied offline and online. In this work, the selection of the set of the features that allowed us to obtain the best results for the classification of this type of signals is presented. In order to compare the results obtained, the Nina PRO DB2 and DB3 databases were used, which contain information on 50 different movements of 40 healthy subjects and 11 amputated subjects, respectively. The sEMG of each subject was acquired through 12 channels in a bipolar configuration. To carry out the classification, a convolutional neural network (CNN) was used and a comparison of four sets of features extracted in the time domain was made, three of which have shown good performance in previous works and one more that was used for the first time to train this type of network. Set one is composed of six features in the time domain (TD1), Set two has 10 features also in the time domain (TD2) including the autoregression model (AR), the third set has two features in the time domain derived from spectral moments (TD-PSD1), and finally, a set of five features also has information on the power spectrum of the signal obtained in the time domain (TD-PSD2). The selected features in each set were organized in four different ways for the formation of the training images. The results obtained show that the set of features TD-PSD2 obtained the best performance for all cases. With the set of features and the formation of images proposed, an increase in the accuracies of the models of 8.16% and 8.56% was obtained for the DB2 and DB3 databases, respectively, compared to the current state of the art that has used these databases.

The implementation of embedded systems for myoelectric signal classification is key to the development of prosthetic and other real-time assistive technologies. This work presents a hand gesture classification system based on an artificial neural network (ANN) deployed on an ESP32 microcontroller. In contrast to approaches that require multiple acquisition channels or complex configurations, the proposed system uses a single surface electromyography (sEMG) channel, reducing cost and complexity. Data were collected from eight participants performing eight hand and wrist gestures. The system achieved classification accuracies of up to 93.05% for six gestures, including rest. Furthermore, the ESP32 implementation demonstrated inference times compatible with real-time operation, achieving an average inference time of 0.17 ms. When using a majority voting strategy, classification accuracy increased to 99%, confirming the feasibility of implementing efficient and accurate gesture recognition on low-cost embedded hardware. These results support its integration into portable and autonomous systems for prosthetic control and other interactive applications.

Hyperspectral imaging (HSI) systems have been demonstrated as a powerful imaging technique due to their high spectral resolution. HSI can obtain the spectrum for each pixel in the image of a scene, a feature that can be exploited to design optical systems with the purpose of analyzing and characterizing objects and identifying processes within the visible electromagnetic spectrum (bandwidth). In this paper, we present an HSI system comprising a diffraction grating placed in the exit pupil of our optical configuration. The spectrum for each pixel associated with the object appears in the first order of diffraction. We used this system to characterize and tune the required spectral band of the image of the captured object obtaining more information than with an optical imaging system. Accordingly, the proposed optical system is suitable to obtain spectral and hyperspectral imaging at low cost compared to an acousto-optic system or other HSI. The scanning system captures hundreds of spectral images associated with the object, obtaining a maximum spectral resolution of 0.26nm or 260 pm for one of our configurations.