China
Africa
Explore chapters and articles related to this topic
Signal Processing in the Era of Biomedical Big Data
Published in Ervin Sejdić, Tiago H. Falk, Signal Processing and Machine Learning for Biomedical Big Data, 2018
Data compression, in turn, is the process of reducing the size of a data file and can be classified as lossless or lossy. Lossless compression reduces data file size by identifying and eliminating statistical redundancy, and hence does not lose any information. Lossy compression, on the other hand, reduces file size by removing unnecessary or less important information, and thus allows for a trade-off between distortion and compression ratio. Within the health care domain, data compression has played a crucial role in medical imaging [23] and in the transmission of biometric data collected via wearables and smart devices [24].
Basics of Video Compression and Motion Analysis
Published in Maheshkumar H. Kolekar, Intelligent Video Surveillance Systems, 2018
Lossless compression reduces the size of data without the loss of any information, and when reconstructed, restores it to its original form. GIF is an example of lossless images compression, which is generally used for text or spreadsheet files, where losing words or financial data could create a problem. For example, if the picture contains 100 pixels of blue color, it can be compressed by saying 100 bluepixels instead of saying 100 times bluepixel. Huffman-coding and run-length encoding exploit redundencies and allow high compression ratios without loss of any data.
Compressive Sensing for Wireless Sensor Networks
Published in Fei Hu, Qi Hao, Intelligent Sensor Networks, 2012
Mohammadreza Mahmudimanesh, Abdelmajid Khelil, Neeraj Suri
In general, we see a variety of compression and decompression requirements in sensory systems, especially distributed sensor networks like WSNs. Compression can be either lossless or lossy. Lossless compression refers to compression techniques that conserve the whole raw data without losing accuracy. Lossless compression is usually used in a common data compression technique which is widely used in commercial compression applications. Some of the most commonly used lossless compression algorithms are run-length encoding, Lempel–Ziv–Welch (LZW), and Lempel–Ziv–Markov chain algorithm. These algorithms are used in many compression applications and several file formats like graphics interchange format, portable network graphics, ZIP, GNU ZIP, etc.Lossy compression, on the other hand, allows unimportant data to be lost for the sake of more important data. Most of the lossy compression techniques involve transforming the raw data to some other domain like frequency domain or wavelets domain. Therefore, such techniques are also called transform coding. Since decades, it has been known that natural signals like audio, images, video and signals recorded from other physical phenomena like seismic vibrations, radiation level, etc. have a sparse or compressible support in Fourier, DCT, and wavelets. An encoder of such natural signals can transform the raw data using a suitable transformation and throw out negligible values. A decoder can recover the original signal with some small error from fewer number of compressed data items.
Realization of RFID-Based DAS Using an RF Transceiver and Huffman Coding
Published in IETE Journal of Research, 2020
Soumen Ghosh, Palash Kumar Kundu
The task of compression consists of two components, an algorithm that takes a message and generates a “compressed” representation is called encoding, and an algorithm that reconstructs the original message or some approximation of it from the compressed representation is called decoding. The data compression is classified as lossy and lossless compression algorithm [15]. The lossy compression reduces the file size by eliminating unnecessary data not recognized by a human being after decoding. Lossless compression handles each bit of data inside the file to reduce the size without losing any data after decoding [16]. In data acquisition, sensor output is collected by the tag and sent to the reader and stored in the memory of the computer. It creates a problem of storage space because large quqntities of data are received by the tag. The compression technique plays a vital role here. The RFID data compression problem is to change the structure of the input stream into an output stream with a reduced data size but with no loss of information [6]. The Huffman coding is a lossless compression technique. Huffman AlgorithmC is set of n character. Q is minimum priority queue. Step 1: n ← |C| Step 2: Q ← C Step 3: for i 1 to n - 1 Step 4: do allocate a new node z Step 5: left[z] ← x ← EXTRACT-MIN (Q) Step 6: right[z] ← y ← EXTRACT-MIN (Q) Step 7: f [z] ← f [x] + f [y] Step 8: INSERT(Q, z) Step 9: return EXTRACT-MIN(Q)
Kinematics-enabled lossless compression of freeway and arterial vehicle trajectories
Published in Journal of Intelligent Transportation Systems, 2019
A consequence of using parsimonious encoding is that interpreting the compressed file is possible only by computer and only because the specific pattern of bit lengths is stored in the file, using the BCD format described earlier. This technique was also widely employed in Lovell (2018).Frequency-based encoding. There are columns of data in the data set that consist of integers that change very seldom through the file. For example, there are very few lanes on a highway. Each new vehicle starts in some given lane and likely changes lanes seldom throughout the file. Rather than repeating the same lane number for each trajectory entry, it saves a lot of space to record only the values and locations of transitions in the data. This is one of many forms of contextual compression exploited in this paper. It works primarily because the analyst knows, ahead of time, that lane changes occur with much lower frequency than the sampling frequency of the data set.Arithmetic compression. It is often the case that, among a particular group of numbers to be stored, the number of unique entries is much smaller than the set itself. Moreover, the distribution of the number of repetitions of these unique numbers might be distinctly non-uniform. In such a case, a number of extremely powerful and ubiquitous lossless compression algorithms are useful, among them Huffman coding and arithmetic compression. In either case, the set of numbers to be encoded is often called the sequence, and the set of unique numbers within that set is called the alphabet. If one knows the distribution of the alphabet over the sequence exactly, then this is the best situation, and if the alphabet is small enough, it makes sense to store this distribution as part of the compressed file. Since the data files studied in this paper are static, the symbol counts can be known exactly, but if the number of unique symbols is too large, it may not be efficient to store them. If they can be approximated fairly accurately with some function that requires only the storage of a few parameters, then arithmetic compression can still be quite robust. This paper makes extensive use of arithmetic compression, including both exact and approximate symbol distributions. The technique essentially forms a binary fraction from recursive embedding of the interval [0, 1], which is subdivided according to the count distribution. This technique is described excellently in Langdon (1984) and Said (2003).Kinematic motion prediction. A kinematic variable such as position can be interpreted as a time series with strong lag-2 correlation. For example, the velocity of a vehicle at time can be predicted fairly accurately using its locations and :