Abstract: A method and a system for providing a monthly scores, a sleep quality index and coach messages to a user for improving the sleep quality. The method includes sleep and activity record data from a wearable device to infer one month of sleep quality evaluation, comparing the results with previous evaluations to coach the users how to change their behaviors to improve their sleep. A Sleep Quality Index is provided based on the estimation of seven components: Sleep Duration, Habitual Sleep Efficiency, Sleep Latency, Use of Sleep Medication, Subjective Sleep Quality, Sleep Disturbances and Daytime Dysfunction. The method further includes analysis of the trends and fluctuations of the scores and the Sleep Quality Index to provide meaningful and insightful messages aiming to provide feedback to the user about how to improve each scores individually along with the Sleep Quality Index, leading to a better user-specific sleep quality.

Abstract: A method that estimates and corrects the heart rate when the photoplethysmography signal is unreliable by using an exponential model with parameters adjusted with machine learning strategies using the reliable HR from the PPG signal and the value of the barometer signal collected by a wearable device after the end of a workout session.

Abstract: Predict aerobic threshold and anaerobic threshold from data obtained from wearable devices during unsupervised exercise sessions, typical of daily exercise routines. These physiological-based predictions from wearable devices have the potential to improve the general health of a wider audience that may request individualized fitness guidance to train at the OATZ. In short, anthropometric and physiological features, such as gender and heart rate, are collected using wearable devices and used to feed machine learning models to predict the AT and AnT that can be used to define the upper and lower bound, respectively, of the optimal aerobic training zone.

Abstract: A method for predicting heart rate (HR) of a user during a physical activity, comprising: monitoring the user during the physical activity with an wearable electronic device comprising a photoplethysmography (PPG) sensor, an accelerometer, a memory and a processor; wherein the memory is configured to store user demographic data, PPG data and accelerometer data detected by the PPG sensor and the accelerometer, respectively. The method includes computing a PPG heart rate estimation and a PPG quality flag using the PPG data and accelerometer data, wherein the PPG quality flag indicates whether the PPG data is reliable or not; computing an exponential heart rate estimation by applying the PPG heart rate estimation, the PPG quality flag and the user's demographic data on an exponential approximation model; if the PPG quality flag is reliable, outputting the PPG heart rate estimation; otherwise, outputting the exponential heart rate estimation.

Abstract: The present invention refers to a method of training a neural network to detect anomalies in a manufacturing product comprising: obtaining a dataset including multiple manufacturing product images and annotation files with coordinates of predetermined anomalous regions corresponding to the manufacturing product images, respectively; selecting a test query set of images and a support set of images from the dataset multiple to train a deep learning model, the support set of images comprises pairs of reference images and each pair comprises at least one anomalous manufacturing product image and at least one non-anomalous manufacturing product image; inputting the test query set into a deep learning model to rate a similarity score based on a similarity distance between the pairs of reference images of the support set based on the annotation files; and adjusting parameters characterizing the deep learning model through a model based on the similarity score.

Abstract: The present invention incorporates reinforcement learning into a beam sweeping framework to select the appropriate set of beams to transmit reference signals in a predefined time interval for covering an angular region. More specifically, a network node starts a learning process to determine the most appropriate subset of beams from a large set of available beams (codebook) to communicate with an associated network node over a radio channel. The transmitter node acquires knowledge from its interaction with other nodes of the wireless network to perform beam sweeping with reduced signaling overhead and latency. More specifically, other advantages, the invention improves the beam management in higher carrier frequencies.

Abstract: The present invention describes a method, system, and a non-transitory computer readable storage medium to mitigate the side effects of imperfect channel estimation, analog processing imperfections, beamforming misalignment, and other imperfections present in massive MIMO wireless communications systems. Through the present invention, the user will have a better communication experience regarding data rate transmission, spectrum efficiency, and, specially, communication reliability. The core of the present invention relies on how the imperfections present in massive MIMO systems are modeled and faced by adding more robustness to the proposed baseband precoder. In addition, the present invention has the capacity to be implemented in any hybrid hardware architecture, where the number of radiofrequency (RF) chains is lower than the number of antennas.

Abstract: Android has been a constant target of cybercriminals that try to attack one of the most used operating systems, commonly using malicious applications (denominated malwares). The present invention uses a lightweight and non-invasive approach to detect malware in the Android system. The method employs a set of specific-type detectors in which each one performs a multi-stage analysis, based on rules and machine learning techniques, in different phases of the application cycle. The invention includes a process to obtain application's characteristics that does not infringe licenses and terms of use of applications; a more efficient process to classify applications on Android devices that requires less processing power; and, finally, by using different detection phases that employ distinct sets of characteristics obtained from the application, the present invention can outperform the detection performance of techniques of the prior-art and reduce the number of misclassified samples.

Abstract: A method for predicting maximal oxygen uptake (VO2Max) in wearable devices with memory restriction using an ensemble of machine learning algorithms is described. The method is related to the fields of well-being, healthcare and artificial intelligence, and includes a technique that predicts maximal oxygen uptake (VO2Max) in running sessions at different paces using wearable devices with memory restriction. The proposed method requires less than 5 KB of memory to run on a wearable device. Specifically, based on user's profile data (age, gender, height and weight), a set of heart rate (HR) and speed readings of a running session, the method is able to estimate VO2Max.

Abstract: A system and method of detecting food intake events from wearable devices. The system comprises a signal bank to store physiological digital data signals collected via a wearable device comprising at least one sensor to sense at least one physiological parameter of a user. The system comprises a preprocessor module to process the physiological digital data signals stored on the data signal bank and to create a descriptive representation of the sensed at least one physiological parameter. The system also comprises a feature extractor module, with two parallel function modes, comprising features that are automatically learned while others are analytically derived from the descriptive representation. Thus, the feature extractor is configured to select at least one of the modes. The system also comprises a probability estimator module to determine whether a food intake event of the user occurs based on the extracted features.

Abstract: A computer implemented method for classifying an input image for new view synthesis in a 3D visual effect, the input image being used to create an image representation comprising a set of target views based on the input 3D visual effect, each target view having an image size of H×W. The method includes computing an image score st for the set of target views based on a number of non-occluded pixels in the set of target views, wherein the image score indicates suitability of the input image for new view synthesis in a 3D visual effect. A system and a non-transitory computer readable storage medium for performing said method for classifying an input image for new view synthesis in a 3D visual effect.

Abstract: A method to detect the presence of SSL-based attacks over a network where a device is connected to, even if the attack pattern is not previously known (zero-day attacks). The invention enabled be used in any mobile device (e.g., smartphones, tablets, notebooks, wearables, Smart TVs, or IoT devices) that relies on wireless communications as it does not require hardware changes and has low computational overhead. The invention can also be used on any device that connects to the Internet by other means, either through the use of mobile communications (4G/5G) or wired networks.

Abstract: A method for generating meta-learning tasks in order to solve few-shot learning classification. The method enables the homogenization of the difficulty level of each task, so the meta-learning process better converges, and the resulting meta-model better generalizes. For each task, the method controls the distances of the data instances within a given class in the input feature domain based on a reference distribution obtained from a known dataset.

Abstract: A method for performing multiple user multiple input multiple output (MU-MIMO) communication, comprising: computing a digital beamformer matrix with a digital precoder for processing baseband signals; processing the baseband signals with a plurality of radiofrequency (RF) chains; computing an analog beamformer matrix with an analog precoder for processing analog or passband signals. The analog precoder comprises: a plurality of adders connectable to the plurality of antennas; a plurality of three-way switches connectable to the plurality of RF chains and the plurality of adders. One of the three terminals of the three-way switches is directly connected to a RF chain, the other terminal is connected to an inverter, and the last one turns off the signal. Computing an optimization function to obtain a maximization of the system sum-rate with basis on the digital and analog beamforming matrixes, the analog precoder sets a position of the three-way switches.

Abstract: A method and apparatus performing multiple user multiple input multiple output (MU-MIMO) communication. The apparatus comprising a hybrid precoder for channel precoding signals to be transmitted to a plurality of users through user channels. The hybrid precoder comprising a digital precoder configured to process signals digitally by a digital beamformer matrix, a plurality of radiofrequency chains and an analog precoder configured to process the signals analogically. The analog precoder comprises a grouped plurality of antennas to thereby allow building of the antennas' array as an array of several subarrays of antennas with each antenna is connected to a two-way switch, whose terminals are connected to a phase inverter and a multiple-way switch and the other terminal of the phase inverter is connected to the multiple-way switch. The number of terminals of the multiple-way switch is equal to the number of RF chains, so that each antenna subarray is switchable among the RF chains.

Abstract: The present invention refers to removal of redundant information from plenoptic point cloud data, reducing the number of bits needed to represent them and thus making the plenoptic point cloud data more suitable to be transferred through a medium of limited bandwidth. The proposed solution is based on predictive differential coding, using the standard color channel of a point clouds as a reference for plenoptic data, and on the application of transforms for greater data compression.

Abstract: A computer implemented data augmentation method comprising receiving a dataset to be processed and, upon the received dataset being unclassified into classes, performing a clustering algorithm to partition the dataset whereby clusters formed are interpreted as the signal classes. The method further includes forming a sample dataset by gathering, for each class of a plurality of classes, at least two sample signals then applying a discrete Fourier transform (DFT) to each sample signal of the sample dataset. The method includes computing frequency parameters of each sample signal to determine, based on a spectral coherence threshold, frequency bands: relevant bands that characterizes a class. The method further includes injecting random noise in a phase spectrum of the non-relevant frequency bands of each sample signal of the sample dataset, to generate a set of augmented sample signals, and applying an inverse DFT, in each of the generated augmented sample signals.

Abstract: A method of managing filesystem elements between a user device and a network connected to the user device comprising: receiving a request related to a filesystem element operation over the network; performing the filesystem element operation; updating a local metadata file based on the operation; downloading an encrypted network metadata file including data of the filesystem on the network from the network to the user device. The method comprises decrypting the encrypted network metadata file on the user device with a network encryption key; and provided the network metadata stored in the network is newer than the updated local metadata file, merging the network metadata file stored in the network with the updated local metadata file to update the local metadata file. The method comprises encrypting the updated local metadata file on the user device with the network encryption key; and uploading the encrypted updated metadata file to the network.

Abstract: A method to compute a variable number of image planes, which are selected to better represent the scene while reducing the artifacts on produced novel views. This method analyses the structure of the scene by means of a depth map and selects the position in the Z-axis to split the original image into individual layers. The method also determines the number of layers in an adaptive way.

Abstract: A computer implemented pre-processing method for facilitating machine learning signal classes separability by evaluating candidate time-frequency transformations to be applied to a set of input signals to generate less redundant features. The computer implemented pre-processing method includes obtaining M candidates of time-frequency transformation rules for the set of input signals, for the M candidates, employing the obtained transformations to calculate time-frequency matrices for at least one signal of each class within the set of input signals and assessing class separability in feature space through the minimization of a multi-parameter objective function that considers the value of the overlapping between the areas of intra-class and the inter-class distributions of distances between the time-frequency matrices. A system and a non-transitory computer readable storage medium for performing the method for evaluating candidate time-frequency transformations and data-augmentation method effects.