National Technical University of Athens (NTUA)
Τεχνικές και μέθοδοι παρακολούθησης ενεργειακής κατανάλωσης και μοντελοποίηση αλγορίθμων επεξεργασίας βιοϊατρικών σημάτων σε ασύρματα δίκτυα αισθητήρων
Abstract
dc:descriptionThe explosive growth of microelectronics along with the sensor industry and wireless communications led to the development of sophisticated electronic devices which integrate microcontroller capacities, sensory circuits, wireless interfaces, printed antennas and supply units. Wireless sensor networks is considered a rather new technological and scientific area with a significant research interest and thousands of publications for all aspects of this issue. Various research groups have given major contribution and proposed sustainable solutions especially in the specific area of biomedical applications of wireless sensor networks. One of the active research issues in the wireless sensor network scientific field is related to the energy consumption of nodes and the determination of network lifetime expectancy which greatly determines the functionality and the network capacity to respond to asynchronous events monitoring and actuation. Two major fields gather the attention of the scientific community that deals with the optimization of protocol parameters implemented in the wireless sensor network and the development of low power consumption electronics. The efficiency of the proposed solutions is rarely explored in real life applications and is mainly limited to the simulation results for network life expectancy improvement. On the other side there are many research methods applied on virtual networks in simulation stage that have been proved to be difficult in implementation. The Doctoral Thesis analyzes a methodology framework for the theoretical calculation of energy consumption in wireless sensor network nodes and simulation results are provided for the estimation of energy consumption node's subsystem. Furthermore, three novel electronic devices are developed for the verification and validation of theoretical and simulated results. Wireless radio subsystem is mainly responsible for a significant part of energy consumption. In the doctoral Thesis this allegation is verified and techniques are developed that target to the minimization of duty cycle for the specific subsystem. The most efficient way of energy consumption reduction is proved to be the long sleep periods for the radio communications subsystem. This is supported by theoretical tools, simulation and measurement campaign results. Moreover, the reduction of data volume transmitted by the radio communication subsystem is accomplished by the implementation of a signal processing method that is applied on data collected by the sensory subsystem. Focusing on biomedical time series, special interest is given on empirical mode decomposition (EMD). The theoretical background of the method is not substantially developed in literature leading to numerical simulation as a means to explore the effect of the method on collected data. The algorithm that implements empirical mode decomposition is partially studied in terms of number of decomposed signals and computation time. This doctoral thesis fills this gap by proposing a methodology road map for the determination of the computational requirements in order to run the algorithm on biomedical time series acquired by means of the wireless sensor node's sensory subsystem. The acceptance of the Empirical Mode Decomposition as a processing tool is stressed by the large number of publications in diverse areas of signal processing including financial applications, fluid dynamics, ocean engineering, electromagnetic field time series analysis and image analysis. Open issues in the research field of empirical mode decomposition are related to the adaptive data analysis methodology in order to find out the underlying processes. In system identification field, the knowledge of the nonlinear characteristics of a physical process has become a hotspot. The lack of a solid base definition and the limited success of a credible methodology test for the nonlinear system identification based only on observed time series remains an unsolved issue. Furthermore, the prediction problem is a challenging task in terms of determining a spline curve for the extension of data. The advantage of empirical mode decomposition is based on the fact that only the IMFs (decomposed signal originating from the application of EMD on the original time series) need to be predicted and this raises the need to identify the value and location of the signal’s extrema. Moreover, the selection of the optimum parameters in EMD algorithm in order to produce the best IMF set is still under investigation.Number of signals produced by the application of the empirical mode decomposition, called Intrinsic Mode Functions (IMF), is unknown prior to the application of the method. A prediction model for the number of IMFs is developed which is based on the statistical properties of the collected biomedical time series. The prediction model outcome is provided as an upper and lower limit. Validation stage application on Physionet biomedical electrocardiogram time series results in a 98% success prediction of the model. Furthermore, a methodology for the statistical significance of the IMFs in the presence of White Gaussian Noise is provided which facilitates the partial signal reconstruction. Doctoral Thesis concludes with the development of a processing time prediction model for the application of the method on resource limited wireless sensor network nodes which is based on statistical properties of the collected biomedical time series. The contribution of the doctoral thesis on the enrichment of the literature related to EMD is multifold. First, it copes with an a priori prediction of the total number of IMFs produced from time series, based on knowledge and measurements of specific parameters and characteristics of the time series. Secondly, it describes a methodology starting from a monocomponent signal and extending to biomedical signals in order to study the performance of EMD in a generalized way regardless of the time series. Thirdly, a model is proposed for the prediction of the number of IMFs that achieves high accuracy and prediction success compared to other existing models. Fourthly, a prediction model for the total computation time required for the execution of the method on wireless sensor network node level is introduced. The strong point of the proposed prediction model is related to the a priori estimation of computation time for the execution of the EMD algorithm on biomedical signals without the execution of the algorithm body.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2011
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Karagiannis, Alexandros
- Καραγιάννης, Αλέξανδρος
Subjects
dc:subject × 20- Ασύρματα δίκτυα αισθητήρων
- Ενεργειακή κατανάλωση κόμβων αισθητήρων
- Θεωρητικός υπολογισμός, προσομοίωση
- Ηλεκτρονικές διατάξεις μετρήσεων
- Εμπειρική μέθοδος αποσύνθεσης
- Μοντέλο πρόβλεψης πλήθους IMF
- Μοντέλο εκτίμησης υπολογιστικού χρόνου
- IMF
- Wireless sensor networks
- Sensor node energy consumption
- Theoretical calculation, simulation
- Electronic measurement circuits
- Empirical mode decomposition
- Intrinsic mode functions
- Prediction model for the number of IMFs
- Computation time estimation model
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/26109