A new algorithm could allow ten times more capacity in the digital signal

A fast, accurate method of analysing the singularities of digital signals has been developed. Among its outstanding characteristics, it makes it possible to detect and to recognize patterns and structures starting from very little data and even to recognize subtle changes and relevant information from low resolution pictures.

 

 

'We can even identify the ocean currents (right picture) just from a single temperature sea surface image (left)'

If you put aside the mathematical definition of singularities, these could be described as points where an abrupt change takes place. In a hypothetical homogeneous space, singularities are the less regular and most informative points that indicate something is happening. Therefore, in a low-resolution portrait, or even in a blurred picture, one can still recognise the contour lines of an object, or the traits of a face: they are the singularity points and contain the main information of the picture.

The example is given by Antonio Turiel, scientist at the Physical Oceanography Department in the CSIC’s Instituto de Ciencias del Mar. He is working with his team to develop applications for ocean monitoring, mainly satellite oceanography. As a result of this research a new algorithm to analyze singularities within digital signals has been devised.

The method, protected by a worldwide patent, can be applied to identify flow patterns in the oceans. However its uses can go much further: from identification of subtle details in medical image diagnostics to the detection of irregularities in materials. It also makes possible the detection of internal sea waves and ocean currents fluxes, to locate ships in the sea and the reconstruction of images with very little data, , what means it could improve the digital signal compression. “Compared to the current standards, it could allow ten times more capacity with the same quality, although is only a theoretical estimation”, says Turiel.

Turiel explains that ‘from one single satellite image we identified 500-km-long wave fronts propagating towards the Indian Ocean interior, something unreported until now with current methodologies.

“It supposes an outstanding improvement in detection technologies. We can even identify the ocean currents just from a single temperature sea surface image”, Turiel points out. The algorithm identifies the most informative points and therefore detects these marine structures. In the resulting image, what can be seen are the lines of these flow lines. “Normally you would need to use a combination of several altimeter satellites and to process a sequence of data to obtain a result as good as this”.

“Originally we intended to apply this methodology to the detection of tsunamis in open ocean using available satellite data, but satellite sensors do not yet provide enough colour depth for that“, adds Turiel.

The scientists are working on other applications such as speech recognition. In collaboration with a team in the French Institut National de la Recherché en Informatique et en Automatique, INRIA, they are actively working on identifying phonemes. The analysis of wind fields for different applications (meteorology, aerodynamic studies, design of wind fields,…) or the study of quotation price series in the stock market are other applications. In the case of the stock market, the scientists have managed to identify different investment and volatility cycles using only the data of the prices with no need of additional information such as the volume of transactions.

More:  digital.CSIC: Obtaining and monitoring of global oceanic circulation patterns by multifractal analysis of MicroWave Sea Surface Temperature images

 

Contact:

Vicepresidencia Adjunta
de Transferencia del Conocimiento
CSIC Catalunya
Artur Sixto
Tel. 934 00 61 00
This email address is being protected from spambots. You need JavaScript enabled to view it.

 

FaLang translation system by Faboba