Dynamic Alternation of Huffman Codebooks for Sensor Data Compression

Daniel Yunge, Sangyoung Park, Philipp Kindt, Samarjit Chakraborty

Research output: Contribution to journalArticlepeer-review

4 Scopus citations

Abstract

Signal compression is crucial for reducing the amount of communication, and hence power consumption of wireless sensors. Lossless compression techniques, such as Huffman coding, are often used in healthcare applications since they do not compromise the integrity of vital signals. Techniques that adapt to changing signal patterns have been proposed. However, most of them involve significant computation overhead or are too simple to maintain high compression rates under changing signal patterns. In this letter, we propose a technique that makes use of multiple codebooks, which are generated offline based on the signal context. In the applications we study, we observe that the symbols that compose a big variety of signals follow Laplacian distributions in which the spread changes over time. This can be effectively utilized to generate a set of codebooks. Then, appropriate codebooks are selected online depending on the currently measured spread, which ensures high compression efficiency and the adaptability to changing signal patterns. Our experiments on real-world medical datasets show that our approach is computationally very efficient, and exhibits competitive compression rates. Our proposed technique outperforms a state-of-the-art compression algorithm, FAS-LEC, in terms of average data reduction by 4.3%, while consuming a similar amount of energy. Compared to the adaptive Huffman method, which achieves near-optimal compression rates, our results indicate energy savings of 19% due to the reduced computational complexity, while the compression rate is improved by 0.6%.

Original languageEnglish
Article number7947176
Pages (from-to)81-84
Number of pages4
JournalIEEE Embedded Systems Letters
Volume9
Issue number3
DOIs
StatePublished - Sep 2017
Externally publishedYes

Keywords

  • Bluetooth
  • body sensor networks
  • codecs
  • compression algorithms
  • Huffman coding
  • maximum likelihood estimation
  • meteorology
  • telemedicine
  • wearable computers

Fingerprint

Dive into the research topics of 'Dynamic Alternation of Huffman Codebooks for Sensor Data Compression'. Together they form a unique fingerprint.

Cite this