Search

Arm-and-hand tracking by technological means allows gathering data that can be elaborated for determining gesture meaning. To this aim, machine learning (ML) algorithms have been mostly investigated looking for a balance between the highest recognition rate and the lowest recognition time. However, this balance comes mainly from statistical models, which are challenging to interpret. In contrast, we present $\mu C^{1}$ and $\mu C^{2}$ , two geometric model-based approaches to gesture recognition which support the visualization and geometrical interpretation of the recognition process. We compare $\mu C^{1}$ and $\mu C^{2}$ with respect to two classical ML algorithms, k-nearest neighbor (k-NN) and support vector machine (SVM), and two state-of-the-art (SotA) deep learning (DL) models, bidirectional long short-term memory (BiLSTM) and gated recurrent unit (GRU), on an experimental dataset of ten gesture classes from the Italian Sign Language (LIS), each repeated 100 times by five inexperienced non-native signers, and gathered with wearable technology (a sensory glove and inertial measurement units). As a result, we achieve a compromise between high recognition rates ($>90\%$) and low recognition times ($ < 0.1 {\mathrm{ s}}$) that is adequate for human–computer interaction. Moreover, we elaborate on the algorithms’ geometric interpretation based on geometric algebra, which supports some understanding of the recognition process. [ABSTRACT FROM AUTHOR]

Objective: In order to evaluate Parkinson disease patients’ response to therapeutic interventions, sources of information are mainly patient reports and clinicians’ assessment of motor functions. However, these sources can suffer from patient's subjectivity and from inter/intra rater's score variability. Our work aimed at determining the impact of wearable electronics and data analysis in objectifying the effectiveness of levodopa treatment. Methods: Seven motor tasks performed by thirty-six patients were measured by wearable electronics and related data were analyzed. This was at the time of therapy initiation (T0), and repeated after six (T1) and 12 months (T2). Wearable electronics consisted of inertial measurement units each equipped with 3-axis accelerometer and 3-axis gyroscope, while data analysis of ANOVA and Pearson correlation algorithms, in addition to a support vector machine (SVM) classification. Results: According to our findings, levodopa-based therapy alters the patient's conditions in general, ameliorating something (e.g., bradykinesia), leaving unchanged others (e.g., tremor), but with poor correlation to the levodopa dose. Conclusion: A technology-based approach can objectively assess levodopa-based therapy effectiveness. Significance: Novel devices can improve the accuracy of the assessment of motor function, by integrating the clinical evaluation and patient reports. [ABSTRACT FROM AUTHOR]

We realized and tested a novel system aimed at discriminating different hand poses by means of an active actuating and sensing approach realized by converting electromagnetic to mechanical waves (and vice versa) and analyzing the characteristics of the waves after their travel through the bones and cartilage of the hand. The actuating part is realized through a transducer, placed on the dorsal part of the hand, performing electromagnetic-to-mechanical energy conversion. The sensing part is realized through an accelerometer, placed on a finger, performing mechanical-to-electromagnetic energy conversion. The way the sound propagates through the fingers mainly depends on the traveled path, the angles of fingers’ flexion, and the (un)matched conditions with surroundings. Therefore, the investigation of the characteristics of the returned signal (in phase shifts and power spectral densities) can furnish information for discriminating among different finger poses and different densities of touched materials. The system highly performed in repeatability and reproducibility of the measures, well-discriminating among hand poses and among densities of touched materials. [ABSTRACT FROM AUTHOR]

Sign languages (SLs) can aid in improving the communication between hearing impaired and hearing communities, but only a limited number of hearing individuals understand it. SL recognition systems tackle this issue by using sensors to acquire gesture data later converted into spoken or written language by means of machine-learning algorithms. Within this frame, we propose a wearable electronic-based system for achieving automatic signer-independent recognition of Italian SL. The system is used for recognizing ten signs performed by 17 inexperienced signers. The wearable electronics measure hand, arm, and forearm gestures, by means of a sensory glove (equipped with ten flex sensors) and six inertial measurement units (IMUs). We adopt three classifiers: artificial neural network (ANN), $k$ -nearest neighbors, and support vector machine, with ANN outperforming the others with an accuracy of 95.07%. However, the real key and novel aspect of this work concerns exploring the importance of each sensor for the overall classification accuracy, with the aim of determining the minimum number of sensors (and so minimizing the complexity of the wearable system) necessary to guarantee an acceptable accuracy. As a result, we could drop out three of the six IMUs, obtaining an accuracy as high as 93.91%. In addition, this valuable result was obtained using a signer-independent approach, which makes the application closer to a real-life situation. [ABSTRACT FROM AUTHOR]

Objective: The complex nature of Parkinson's disease (PD) makes difficult to rate its severity, mainly based on the visual inspection of motor impairments. Wearable sensors have been demonstrated to help overcoming such a difficulty, by providing objective measures of motor abnormalities. However, up to now, those sensors have been used on advanced PD patients with evident motor impairment. As a novelty, here we report the impact of wearable sensors in the evaluation of motor abnormalities in newly diagnosed, untreated, namely de novo, patients. Methods: A network of wearable sensors was used to measure motor capabilities, in 30 de novo PD patients and 30 healthy subjects, while performing five motor tasks. Measurement data were used to determine motor features useful to highlight impairments and were compared with the corresponding clinical scores. Three classifiers were used to differentiate PD from healthy subjects. Results: Motor features gathered from wearable sensors showed a high degree of significance in discriminating the early untreated de novo PD patients from the healthy subjects, with 95% accuracy. The rates of severity obtained from the measured features are partially in agreement with the clinical scores, with some highlighted, though justified, exceptions. Conclusion: Our findings support the feasibility of adopting wearable sensors in the detection of motor anomalies in early, untreated, PD patients. Significance: This work demonstrates that subtle motor impairments, occurring in de novo patients, can be evidenced by means of wearable sensors, providing clinicians with instrumental tools as suitable supports for early diagnosis, and subsequent management. [ABSTRACT FROM AUTHOR]

Presents a study that introduced a novel input pattern that becomes an optimum trade-off between accuracy and complexity while avoiding the use of a tuner for fast noise measurements. Key elements in active device noise characterization; Advantages offered by the proposed pattern; Results obtained with the proposed impedance pattern.