"Applied Wavelet Technology Inc." team is open for collaboration on data analysis in
aviation, image compression, and medicine.

Recent progress in mathematical analysis of very complicated processes is really very impressive. The completely new method which is called wavelet-analysis has been developed and successfully applied in many fields of science and for practical purposes. Using wavelets one can develop effective computer algorithms for image and audio compression, for pattern recognition and object detection, for analysis of low signals over strong background etc. They can be used to save enormous information volume in small packages as it has been done for fingerprints and face recognition. For example, the storage of fingerprints is greatly simplified because it requires much less computer memory (about 10000 times less). The time consuming is also reduced so much that some unsolvable (due to enormous time and memory consuming) problems become solvable (e.g. instead of 10^100 plane waves one uses 10^7 wavelets). Therefore, these algorithms overperform substantially all previously exploited methods.

Our group is very much experienced in wavelets and in their applications to theoretical physics and some practical problems. We have developed the software for pattern recognition in nuclear collisions, for prediction of electronic properties of solid states, for data compression, for detection of precursors of engine failure, for blood cell classification etc. The elegant program packages have been worked out. Their efficiency is close to the pick-up performance of computers used.

This software modified to any definite tasks can be applied to such extremely important problems as diagnosis and analysis of engine disturbances (e.g. for cars or airplanes), irregularities in gas pipelines or in some mechanical devices, quality of raw metal blocks, data compression, analysis of biological or medicine data etc.

Using our experience, we can propose to work on these or similar practical problems with your company.


Aviation

We are ready to collaborate on tuning our software to specific requirements of consumers and, particularly, on finalizing our project of the wavelet analysis of combustion processes in industrial gas turbines plants, e.g., such as GTX100.

At present, we have the software with a streaming 1-dimensional algorithm which has been successfully used to discover the precursors of stall and surge processes in gas turbines and can be modified to adjust it to other particular problems.

Two patents N 2154813 and N 2149438 dated 19.03.1999 for diagnosing and automatic control of engines have been obtained in Russia and patent applications are submitted in USA and Canada. There are several publications of our group on this topic.

The method.

An individual sample of data, e.g., of pressure variations in a gas turbine compressor, is wavelet transformed and various characteristics of distributions of wavelet coefficients are analyzed. They reveal the location of specific disturbances in the signal. It can be used to control and optimize the engine operation.

We are ready to collaborate on tuning our software to specific requirements of consumers and, particularly, on finalizing our project of the wavelet analysis of combustion processes in industrial gas turbines plants, e.g., such as GTX100.

At present, we have the software with a streaming 1-dimensional algorithm which has been successfully used to discover the precursors of stall and surge processes in gas turbines and can be modified to adjust it to other particular problems.

Two patents N 2154813 and N 2149438 dated 19.03.1999 for diagnosing and automatic control of engines have been obtained in Russia and patent applications are submitted in USA and Canada. There are several publications of our group on this topic.

Performance. The continuous signal recorded in the audio cassette R-124 is transformed into a sampled digitized form by processing it in the 8-channel SONY DAT recorder PC200Ax. The digital data are wavelet transformed and analyzed according to our program packages. In principle, this analysis can be done for engines of any type, and not only for gas turbines.

Besides, we have developed the software for pattern recognition and wavelet compression of 2-dimensional images. The algorithms can be used as highly efficient tools for specially purposed image analysis and compression, e.g., in video and audio compression, in medical imaging analysis or in printing.


Image Compression

We are ready to collaborate on tuning our software to specific requirements of consumers and on finalizing our project of the wavelet compression of 3-dimensional images.

At present, we have:

1. the software for a wavelet compression of 2-dimensional images (demo available);

2. streaming 1-dimensional algorithm which can be adjusted to particular problems;

3. the software for pattern recognition in 2- and 1-dimensional images.

The coding method.

An individual frame (2D-movie image) is wavelet transformed into a sequence of 2D coded frames with "time" direction and compressed according to adaptive streaming wavelet coding.

Estimated performance (using PIII-700 prior to optimization and wavelet tuning).
3 frames/sec for PAL standard (352x288, 24 bits/pixel, 25 f/sec);
4 frames/sec for NTSC standard (352x240, 24 bits/pixel, 30 f/sec).

It can be improved up to a factor 3 if the proper algorithm is prescribed for a given processor, architecture and frame size. To achieve the upper limit, we need an essential support for such kind of improvements to be done.

Estimated bit rates.

The program is adjusted to streams less than 800 kbits/sec for video (NTSC) and 50 kbits/sec for audio streams.

Thus the initial untuned characteristics (performance at a given compression level) are similar to MPEG4 standard. However, our algorithms provide better quality for the smooth tone transitions (no JPEG-MPEG squares) as well as better general quality at the same compression level or much stronger compression for "low" quality streams (like Internet video conferences with low bit rate). Therefore, the "practical quality" of the image is substantially higher for these algorithms as compared to JPEG-MPEG. These algorithms can also be used as highly efficient tools for specially purposed image compression, e.g., in medical imaging analysis or in printing, as well as for diagnosis and analysis of engine disturbances etc.


Medicine

We are ready to collaborate on tuning our software to specific requirements of consumers, on finalizing our projects of analysis of blood cells, of EEG and of spermatozoa, as well as on other proposals of pattern recognition and analysis of medical objects.

At present, we have:

1. the software for automatic classification of erythrocytes;

2. the software for a wavelet compression of 2-dimensional images (demo available);

3. streaming 1-dimensional algorithm which can be adjusted to particular problems;

4. the software for pattern recognition in any 2- and 1-dimensional images.

The set of 1-dimensional images is analyzed, e.g., in EEG by wavelet decomposition and automatic detection of the irregularities of a signal. The automatic pattern recognition of 2-dimensional images is done for blood cell (e.g., erythrocytes classification) or spermatozoa analysis. The elaborated package of programs for detection of blood cell or spermatozoon contours and subsequent image analysis has been used.

The coding method.

An individual frame (2D-movie image) is wavelet transformed into a sequence of 2D coded frames with "time" direction and compressed according to adaptive streaming wavelet coding.

Performance. The automatic wavelet analysis of blood cells reduces substantially the time required for analysis of a blood sample and avoids human factors which could lead to some mistakes. The program package using wavelet algorithms has been attached to traditional devices for blood cell analysis in several Moscow clinics and strongly impoved their characteristics. It has been approved by medical doctors. It can be used also in cytological, histological, parasitological etc studies. The "practical quality" of the image is substantially higher for these algorithms as compared to JPEG-MPEG. Therefore, these algorithms can also be used as highly efficient tools for specially purposed image compression, e.g., for transfer of medical images to different destinations.