Performance evaluation of Wearable Computing Frameworks
1. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
16o WPerformance
Performance Evaluation of Wearable Computing
Frameworks
Michelle Cacais1, Danielo G. Gomes1, Paulo Armando1 and Rossana Andrade1
1Grupo de Redes de Computadores, Engenharia de Software e Sistemas (GREat)
Universidade Federal do Ceará (UFC)
2 - 6 de julho de 2017
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 1 / 20
2. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Summary
1 Introduction
2 Wearable Computing
3 Material and methods
Selected frameworks
Case study
4 Performance evaluation
Experimental Setup
Results
5 Conclusion
6 References
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 2 / 20
3. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Introduction
Wearable Computing is one of the IoT core technologies
However, developing artifacts using wearable technology is a difficult task
As a consequence, a wearable device becomes complex and hard to reuse
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 3 / 20
4. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Introduction
Research and development in wearable computing have brought functional and
attractive solutions to users
Figure: Oculus Rift
Fonte:
https://www3.oculus.com/en-
us/rift/
Figure: Fitbit Flex
Fonte:
https://www.fitbit.com/flex
Figure: Smartwatches
Fonte:
https://www.android.com/wear/
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 4 / 20
5. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Introduction
Here we propose a systematic performance evaluation of three selected wearable
computing frameworks
We used metrics such as energy consumption, memory footprint and
accuracy
We take into account some qualitative aspects such as easy of development
and debugging
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 5 / 20
6. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Wearable Computing
Figure: A generic overview of Wearable Computing
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 6 / 20
7. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Challenges in Wearable Computing
Energy consumption: wearables face a worst battery challenge when compared
to phones, because of the smaller sizes
Data accuracy: development of wearable infrastructures involves different
hardware, what means dissimilar accuracy
Memory footprint: body sensors are rich sources of data, so the storage system
has to be capable to support it
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 7 / 20
8. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
How we selected the frameworks
We made a sort of simple systematic review using search strings as
Wearable computing framework
Wearable computing rapid development
Body sensor networks framework
Also, It has been taken into account availability of the code, freshness and
relevance
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 8 / 20
9. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Selected frameworks
SPINE
Figure: Signal Processing In Node Environment (SPINE)
Source: http://spine.deis.unical.it/spine.html
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 9 / 20
10. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Selected frameworks
BNSM
Figure: Body Network Service Manager (BNSM)
Source: adapted from Iyengar (2009), p. 16
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 10 / 20
11. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Selected frameworks
IBM Bluemix
Figure: Body Network Service Manager (BNSM)
Source: http://www.ibm.com/developerworks/library/ba-bluemix-wearables/index.html
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 11 / 20
12. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Case study
Case study
In this case study, each framework have been used to design an application to
recognize and classify human movement
Users became capable to communicate with devices with no need of touch or
speech
This is specially interesting for Wearable Computing
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 12 / 20
13. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Experimental Setup
System parameters
The system parameters include a smartphone Samsung J1, with a textbf1230
mAh battery, a processing power of Dual-Core 1.2 GHz and operating system
Android 5.0 (Lollipop)
The server side was implemented in Java Language for both SPINE and BNSM
For IBM Bluemix, it was used the cloud of the company
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 13 / 20
14. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Experimental Setup
Workload parameters
Quantity of data: 20 samples for each movement (80 in total)
Number of possible movements: 4
Necessary time to complete one movement: 35 seconds
Total time of execution: 12 minutes
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 14 / 20
15. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Results
Energy consumption
Figure: Energy consumption
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 15 / 20
16. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Results
Memory footprint
Figure: Memory footprint
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 16 / 20
17. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Results
Accuracy
Table: Accuracy of the data
Sitting Standing Walking Falling
SPINE 95% 93% 93% 99%
BNSM 95% 95% 96% 100%
IBM 96% 97% 98% 100%
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 17 / 20
18. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Results
Overview of results
Table: Average of the results with a confidence interval
Energy consumption (mW/s) Memory footprint (KB) Accuracy (%)
SPINE 6.9 ± 1.7 4.2 ± 0.8 93 ± 1
BNSM 4.8 ± 1.7 4.5 ± 0.8 95 ± 1
IBM 5.8 ± 1.7 3.2 ± 0.8 97 ± 1
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 18 / 20
19. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
Conclusion
All of the frameworks help with heterogeneity of protocols and nodes
The solution provided by IBM Bluemix was the easiest to use for rapid prototyping
SPINE and BNSM have shown similar results, but the second one was more
efficient for energy saving
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 19 / 20
20. Introduction Wearable Computing Material and methods Performance evaluation Conclusion References
References
BELLIFEMINE, Fabio et al. SPINE: a domain-specific framework for rapid prototyping
of WBSN applications. Software: Practice and Experience, v. 41, n. 3, p. 237-265,
2011.
BORGIA, Eleonora et al. Special issue on “Internet of Things: Research challenges
and Solutions”. Computer Communications, v. 89, p. 1-4, 2016.
GHEITH, Ahmed et al. IBM bluemix mobile cloud services. IBM Journal of Research
and Development, v. 60, n. 2-3, p. 7: 1-7: 12, 2016.
HUSSEIN, Aziza I. Wearable computing: Challenges of implementation and its future.
In: 12th Learning and Technology Conference, 2015. 35752 2015. IEEE, 2015. p.
14-19.
IYENGAR, Sameer et al. A framework for creating healthcare monitoring applications
using wireless body sensor networks. In: Proceedings of the ICST 3rd international
conference on Body area networks. ICST (Institute for Computer Sciences,
Social-Informatics and Telecommunications Engineering), 2008. p. 8.
Michelle Cacais, Danielo G. Gomes, Paulo Armando and Rossana Andrade UFCPerformance Evaluation of Wearable Computing Frameworks 2 - 6 de julho de 2017 20 / 20