Activity recognition models based on wearable devices are becoming increasingly popular. However, models that are trained and tested on the same players show a large bias and are not generalizable to previously unseen players. In this paper, we tackle the badminton stroke recognition problem from this perspective, comparing the performance of individual and generalized models based on an accelerometer and a gyroscope, and identifying which components of the solution can maximize the performance of generalized models. First, we describe a simple convolutional neural network trained to classify 7 types of stroke. Second, the model is extended in a hybrid way to identify two additional classes (movement and rest). Third, data augmentation is applied on the training set. Fourth, transfer learning is applied to use data from the test player to fine-tune the generalized model and attempt to reach the performance of an individual model. These models are evaluated on a dataset collected from amateur players, both in a controlled environment and in a match simulation. The results showed a large difference between the performance of individual and generalized models; however, the latter could be improved by increasing the number of players in the training set, by data augmentation, and by transfer learning, highlighting the necessity of larger datasets in this field.

Fingerprint classification is a frequent approach to deal with very large scale databases in fingerprint recognition. In the last few years, several proposals based on Convolutional Neural Networks have pushed state of the art results even further. However, it has also been proven that such networks are prone to be overconfident in the predictions of the classes, which may have an impact on their performance. This paper aims to study the problem from a systematic point of view. First, it is determined that the most common network to classify fingerprints does suffer from badly calibrated predictions. Second, two calibration methods (temperature scaling and Dirichlet calibration) are applied to correct for this tendency. Third, a modified search strategy is proposed, which makes use of the calibrated class probabilities predicted by the classifier to further reduce the penetration rate and avoid the negative impact of impostor input fingerprints. Fourth, all the proposals are evaluated on five datasets, which combine synthetic and real fingerprints of different qualities. Dirichlet calibration led to improved predicted class probabilities, which in turn allowed for further reduction of the penetration, while maintaining a good trade-off with respect to the false rejection rate.

Fingerprint classification is one of the most common approaches to accelerate the identification in large databases of fingerprints. Fingerprints are grouped into disjoint classes, so that an input fingerprint is compared only with those belonging to the predicted class, reducing the penetration rate of the search. The classification procedure usually starts by the extraction of features from the fingerprint image, frequently based on visual characteristics. In this work, we propose an approach to fingerprint classification using convolutional neural networks, which avoid the necessity of an explicit feature extraction process by incorporating the image processing within the training of the classifier. Furthermore, such an approach is able to predict a class even for low-quality fingerprints that are rejected by commonly used algorithms, such as FingerCode. The study gives special importance to the robustness of the classification for different impressions of the same fingerprint, aiming to minimize the penetration in the database. In our experiments, convolutional neural networks yielded better accuracy and penetration rate than state-of-the-art classifiers based on explicit feature extraction. The tested networks also improved on the runtime, as a result of the joint optimization of both feature extraction and classification.