1. 1 Compression and Storage Schemes in a Sensor Network with Spatial and Temporal Coding Techniques You-Chiun Wang, Yao-Yu Hsieh, and Yu-Chee Tseng IEEE Vehicular Technology Conference, 2008

2. 2 Outline Introduction Multi-resolution compression and storage (MCS) framework Compression and storage schemes Implementation and experimental results Conclusions

3. 3 Introduction The communication overhead will dominate sensor node’s energy consumption Sensing data reported from sensor nodes often exhibit a certain degree of data correlation  Spatial correlation  Temporal correlation People may query different resolutions of sensing data from a wireless sensor network

4. 4 Multi-resolution compression and storage (MCS) framework

5. 5 Compression and storage schemes Spatial compression scheme Temporal compression scheme Storage scheme

6. 6 Spatial compression scheme Layer-1 compression Layer-i (i > 1) compression Decompression Compression ratio: ( 0 ≦ γ ＜ 1 )

7. 7 Layer-1 compression A layer-1 processing node collects the sensing data from the sensor nodes in its block M = (s i,j ) k×k M = 2827 28 29 29 28 28 29 30 29 28 29 29 29 28 28

8. 8 Layer-1 compression (2D-DCT) Two-dimensional discrete cosine transform (2D-DCT) method 2D-DCT will compact those significant values in the upper-left part of the transformed matrix

9. 9 Layer-1 compression (RZS) A reduced zigzag scan (RZS) method is applied to translate M’ into an one dimensional array k 2 ×λ λ = 1 −γ

10. 10 Layer-i (i > 1) compression Reduce the length of array D (passed from the layer i−1) to λ i × k 2 elements Layer-1 Layer-2

11. 11 Decompression The sink recovers the corresponding array D to a two-dimensional matrix M’ = (t i,j ) k×k Adopt the inverse 2D-DCT method to transform M to a new matrix M’’ = (s i,j ) k×k

12. 12 Temporal compression scheme The temporal compression scheme is performed by each sensor node Users can specify a small update threshold δ to determine whether a node should transmit its data or not δ= 2°C S 1,1 = 28°C Range: 28°C ± 2°C

13. 13 Storage scheme(1/2) For a node i, we will store frames f t, f t−1, f t−3, f t−7, · · ·, and f t−2 ni−1 +1 123 ftft f t−1 f t−3 431 542 → →

14. 14 Storage scheme(2/2) f j has been stored in node i’s local memory, node I directly replies f j to the sink j < t−2 n i −1 +1, node i replies a fail message to the sink because f j is too old to be stored in node i 542 f 3 = ? (f 4 +f 2 )/2

15. 15 Implementation and experimental results We use the MICAz Motes as sensor nodes and processing nodes Set the system parameters α = 4 and k = 2 We use this prototype to collect indoor temperatures during 25 hours The compression ratio γ is set to 0.25 The update threshold δ is set to 0.2°C

16. 16 The total amount of message transmissions

17. 17 Average temperatures reported by the 16 nodes

18. 18 Conclusions MCS provides multi-resolution data compression and storage in a wireless sensor network MCS can effectively reduce message transmissions of sensor nodes MCS framework not only significantly reduces the message transmissions but also preserves important characteristics of sensing reports

19. 19 Thank you!

20. 20 M 262830 262829 262728 M’ 77.9118.5126.7 110167.4172.9 109.8161166.5