We have recently developed a next version of our multimodal palm biometric system prototype. This system can capture palm vein, palmprint, and geometrical information in non-contact way. We have also included an interactive touchscreen interface for demonstration purposes as well as to guide the person how to place his palm in front of the reading device. For demonstration purposes we have developed interactive touchscreen interface, where people can choose between image acquisition in infrared light or visible light spectrum, as well as to explore their filtered image of the vein pattern. Camera works at the frame rate of 30 fps. Filtering of an image is also performed in the real-time (30 fps). System prototype was demonstrated at the BIOSIG 2013 conference along the paper “FPGA based palmprint and palm vein biometric system” where it received best poster award.

Recently we have developed a palm print and palm vein biometric prototyping system. Main task for such prototyping system is to demonstrate various ways how a palm based biometric system can be implemented. Two biometric parameters (palm prints and palm veins) can be used separately or together. Smart cards can be used to develop biometric system that stores biometric data on card of performs match on card comparison. For intuitive interface with the system an LCD touch screen display is added.

(more…)

To extract palm vein pattern and palm print structure from acquired images, it is necessary to apply a filtering and feature extraction algorithm. Therefore, we have implemented complex 2D matched filter method in FPGA. Input image from the camera module is filtered “on the fly” without saving data in SRAM IC. Only the RAM bits that are available in FPGA are used. Filtering is performed in the real time with 60 fps and a resolution of 752 x 480. After the filtering with CMF the most significant vectors from different regions of the palm are extracted for further processing. Images below show the input image, image filtered with CMF and the vector set that is extracted from the CMF result.

Result of CMF on FPGA

(more…)

To ease the debugging process of software design a USB communication interface for FPGA DE2 development board has been developed. Asynchronous data transfer using FT2232 FTDI IC were chosen. With developed data interface it is possible to send data with speeds up to 5 MB/s. Also a software for easy PC to FPGA communication was developed. Currently it is possible to trigger the capture of an image, save image to PC, and send data to FPGA.

Multimodal biometric systems use two or more biometric parameters to increase the overall system’s performance and precision. As we use palm as out main biometric object, it is important to acquire images of a palm in two different light spectra. First we can capture image in visible light to obtain image of a palm print structure and second we can capture image in infrared light spectra  to capture palm vein structure. Therefore, an image acquisition design in FPGA is developed, to acquire images in infrared and visible light spectrum from previously developed camera module. Data is stored in SRAM and can be displayed on VGA screen, on the external LCD display module or can be sent to a PC by using a USB data interface that is also developed for easy software debugging and data transfer between FPGA and PC . Few of the acquired sample images are shown below:

Palm capture (infrared light)

(more…)

Palm ROI detection algorithm has been developed to process images from CASIA multi-spectral palmprint database that can be found in Biometrics Ideal Test homepage: http://www.idealtest.org/dbDetailForUser.do?id=6

(more…)

Previously we have described Complex 2D Matched Filter (CMF) that is an angle invariant line-like object detection filter. Then we introduced Non Halo – Complex 2D Matched (NH-CMF) filter that gets rid of the unwanted halo artifacts in the filtering process.  Those filters were designed to extract only lines from images. What we have discovered is that we can generalize the CMF filter for different line-like object detection. Therefore we can detect object boundaries, lines, and different line crossings. This filter is called Generalize Complex 2D Matched Filter (GCMF). The objects that can be extracted from the images by using first 5 orders of the GCMF filter are shown below.

  (more…)

Part of the project is to integrate a smart card system with overall biometric system. At first, raw biometric data comparison software for Java card was developed and then the biohash’s biocode comparison algorithm implemented in Java card. The problem to develope an algorithm for smart cards is that they don’t support floating point numbers. That means that the algorithms must be optimized for integer computations. Our smart cards support 16 bit signed integers, that means that there will be computational errors because of the low precision. The cards are programmed by using a subset of java. Below is a table where the computational errors and running time can be seen. It can be seen that comparing the encrypted data, the computation error is about 2 times smaller. That is because the operations are straight forward and only needs to compare 2 series of bits.

(more…)

Previously described Complex 2D Matched Filter (CMF) is an angle invariant line-like object detection filter. CMF provides additional information about angular orientation of extracted obejcts, which simplifies the segmentation procedure. Although CMF was successfully used for palm vein extraction from images CMF produces unwanted halo artifacts around the extracted objects (in our case – veins). Therefore, we have introduced an improved CMF approach that gets rid of the unwanted artifacts. This method is called Non Halo – Complex 2D Matched Filter (NH-CMF). Below is an example of the palm image filtered with CMF and NH-CMF. For further information please see the paper on NH-CMF in publications page.

Palm image filtered with CMF

Palm image filtered with NH-CMF

A part of the project is dedicated for the security of biometric data. For this purpose we’ve implemented the biohash algorithm to encrypt palm vein and palm print data. First biometric data is processed with CMF to obtain the feature vectors that represent the palm print or vein structure. From this information we create a data vector (of size n) that consists of feature vector descriptors.

(more…)