Motor Unit Spike Trains Recontruction from sEMG Signals for Gesture Classification on a Low-Power Processing Platform

Hand gesture recognition based on surface electromyography (sEMG) signals is a promising approach for the development of intuitive human-machine interfaces (HMIs) in domains such as robotics and prosthetics. The sEMG signal arises from the muscles' electrical activity, and can thus be used to recognize hand gestures. The decoding from sEMG signals to actual control signals is non-trivial; typically, control systems map sEMG patterns into a set of gestures using machine learning, failing to incorporate any physiological insight. This master thesis aims at developing a bio-inspired hand gesture recognition system based on neuromuscular spike extraction rather than on simple pattern recognition. The system relies on a decomposition algorithm based on independent component analysis (ICA) that decomposes the sEMG signal into its constituent motor unit spike trains, which are then forwarded to a machine learning classifier. Since ICA does not guarantee a consistent motor unit ordering across different sessions, 3 approaches are proposed: 2 ordering criteria based on firing rate and negative entropy, and a re-calibration approach that allows the decomposition model to retain information about previous sessions. Using a multilayer perceptron (MLP), the latter approach results in an accuracy up to 99.4% in a 1-subject, 1-degree of freedom scenario. Afterwards, the decomposition and classification pipeline for inference is parallelized and profiled on the PULP platform, achieving a latency < 50 ms and an energy consumption < 1 mJ. Both the classification models tested (a support vector machine and a lightweight MLP) yielded an accuracy > 92% in a 1-subject, 5-classes (4 gestures and rest) scenario. These results prove that the proposed system is suitable for real-time execution on embedded platforms and also capable of matching the accuracy of state-of-the-art approaches, while also giving some physiological insight on the neuromuscular spikes underlying the sEMG.

Effects of flooding on coastal vegetation and salt marshes

This thesis examines the effects of flooding on coastal and salt marsh vegetation. I conducted a field experiment in Bellocchio Lagoon to test the effects of different inundation periods (Level 1 = 0.468 or 11.23 hours; Level 2 = 0.351 or 8.42 hours; Level 3 = 0.263 or 6.312 hours; Level 4 = 0.155 or 3.72 hours; Level 5 = 0.082 or 1.963 hours; Level 6 = 0.04 or 0.96 hours) on the growth responses and survival of the salt marsh grass Spartina maritima in summer 2011 and 2012. S. maritima grew better at intermediate inundation times (0,351 hours; 0,263 hours, 0,115 hours; 0,082 hours), while growth and survival were reduced at greater inundation periods (0,468 hours). The differences between the 2011 and 2012 experiment were mainly related to differences in the initial number of shoots (1 and 5, respectively in 2011 and 2012). In the 2011 experiment a significant lower number of plants was present in the levels 1 and 6, the rhizomes reached the max pick in level 4, weights was major in level 4, spike length reached the pick in level 3 while leaf length in level 2. In the 2012 experiment the plants in level 6 all died, the rhizomes were more present in level 3, weights was major in level 3, spike length reached the pick in level 3, as well as leaf length. I also conducted a laboratory experiment which was designed to test the effects of 5 different inundation periods (0 control, 8, 24, 48, 96 hours) on the survival of three coastal vegetation species Agrostis stolonifera, Trifolium repens and Hippopae rhamnoides in summer 2012. The same laboratory experiment was repeated in the Netherlands. In Italy, H. rhamnoides showed a great survival in the controls, a variable performance in the other treatments and a clear decrease in treatment 4. Conversely T. repens and A. stolonifera only survive in the control. In the Netherlands experiment there was a greater variability responses for each species, still at the end of the experiment survival was significantly smaller in treatment 4 (96 h of seawater inundation) for all the three species. The results suggest that increased flooding can affect negatively the survival of both saltmarsh and coastal plants, limiting root system extension and leaf growth. Flooding effect could lead to further decline and fragmentation of the saltmarshes and coastal vegetation, thereby reducing recovery (and thus resilience) of these systems once disturbed. These effects could be amplified by interactions with other co-occurring human impacts in these systems, and it is therefore necessary to identify management options that increase the resilience of these systems.

Individuazione dei campi recettori di cellule in corteccia visiva primaria (v1), attraverso lo studio di dati ad alta dimensionalita

Il progetto di ricerca che ho svolto in questi mesi si è focalizzato sull'integrazione dei risultati raggiunti grazie all'elaborazione di nuovi dati sperimentali. Questi sono stati prelevati dalla corteccia visiva di macachi, attraverso l'utilizzo di tecniche di registrazione elettro-fisiologiche mediante array di micro-elettrodi[25], durante la presentazionedi alcuni filmati (sequenze di immagini o frames). Attraverso la tecnica del clustering, dalle registrazioni degli esperimenti sono stati raggruppati gli spike appartenenti ad uno stesso neurone, sfruttando alcune caratteristiche come la forma del potenziale d'azione. Da questa elaborazione e stato possibile risalire a quali stimoli hanno prodotto una risposta neurale. I dati messi a disposizione da Ringach non potevano essere trattati direttamente con le tecniche della spike-triggered average e della spike-triggered covariance a causa di alcune loro caratteristiche. Utilizzando filtri di Gabor bidimensionali e l'energia di orientazione e stato pero possibile modellare la risposta di cellule complesse in corteccia visiva primaria. Applicare questi modelli su dati ad alta dimensionalita immagini molto grandi), sfruttando la tecnica di standardizzazione (Z-score), ha permesso di individuare la regione, la scala e l'orientazione all'interno del piano immagine dei profili recettivi delle cellule di cui era stata registrata l'attività neurale. Ritagliare tale regione e applicare la spike-triggered covariance su dati della giusta dimensionalita, permetterebbe di risalire ai profili recettivi delle cellule eccitate in un preciso momento, da una specifica immagine e ad una precisa scala e orientazione. Se queste ipotesi venissero confermate si potrebbe marcare e rafforzare la bontà del modello utilizzato per le cellule complesse in V1 e comprendere al meglio come avviene l'elaborazione delle immagini.