Abstract: TMS320VC5409 isⅡThe company promotes section of digital signal processor (DSP), has the good performance price ratio. Introduces one kind based on this DSP fingerprint recognition system. Gives its design proposal, the fingerprint recognition algorithm design mentality hardware architecture, key discussion to fingerprint pretreatment algorithm improvement, TMS320VC5409 and F37S200 characteristic and CPLD logical design.
Key words: Fingerprint recognition system; Recognition algorithm; Digital signal processor; Complex programmable logical component
1 introduction
Along with the computer and information technology’s unceasing development, the biological features recognition engineering research has received the widespread attention. The biological features recognition is differentiates the biological individual with organism’s biological features the computer technology, it studies the biological features including the face shape, the fingerprint, the hand lines on the hand, the iris, the retina, the sound, the build, are familiar with and so on personally, corresponding recognition technology including person face recognition, fingerprint recognition, lines on the hand recognition, iris recognition, retina recognition, speech recognition, build recognition, keyboard rap recognition, signature recognition and so on. This article introduced the fingerprint recognition is the biological recognition technology one kind, its relative other recognition system (for example iris, face shape) has many superiority. The first fingerprint has the distinctive quality and the stability, and the signal easy to gather, the recognition algorithm was very already mature. At present, carried on the fingerprint recognition based on general PC machine the technology already to start to promote massively. But its construction cost is high, the system is complex, the volume is huge. Can only use in the fixed place, obviously does not suit in systems and so on security, network, handset, entrance guard applies. Enhances and the price large scale drop unceasingly along with the DSP performance, its application scope also unceasingly expands. And, the important application domain is the real-time signal processing. The fingerprint recognition system is high to the speed request, the computation load is big, in treating processes’ many operations are the multiplication and the additive operation, this is precisely the DSP strong point.
2 fingerprint recognition system’s principle and composition
The fingerprint recognition technology use takes likely the equipment read fingerprint image, through the recognition software extraction fingerprint image’s in characteristic data, then the result distinction fingerprint all person status which obtains according to the matching algorithm.
The fingerprint recognition system is composed of the off-line and the online two parts, mainly involves fingerprint processes and so on image gathering, image pretreatment, feature extraction, characteristic match. In system’s off-line part, gathers the meter gathering fingerprint with the fingerprint. Withdraws the characteristic point, then preserves the characteristic point in the database. Forms the fingerprint template storehouse. In system’s online part, gathers the meter gathering fingerprint with the fingerprint. Withdraws the characteristic point, then and the preservation carries on these details in the database template characteristic point the match, the judgment input characteristic point and the template characteristic point whether to come from the identical finger’s fingerprint. Automatic fingerprint recognition system diagram as shown in Figure 1.
3 software designs
3.1 image pretreatments
Because receives each kind of factor the influence, gathers the fingerprint image is one contains the noise many gradation images. Needs the image to carry on the pretreatment regarding this first. The fingerprint image pretreatment is refers to low quality fingerprint image basis certain algorithm which and the treating processes gathers causes its grain line structure clear. As far as possible prominent and the retention inherent characteristic information avoids having the false characteristic information. The fingerprint image pretreatment is in the fingerprint automatic diagnosis process first step, its quality immediate influence fingerprint automatic diagnosis system’s effect. Image pretreatment including filter, binaryzation, refinement and denoising.
Because in the fingerprint chart the grain line has the directivity in the partial scope to be consistent, width basic equal, spacing basic same level characteristic, therefore improves based on fingerprint these characteristics to the traditional algorithm. Uses the directional diagram filter to come to the fingerprint image to carry on processing. The first fingerprint grain line’s direction in O~π between the average division is K fundamental direction, each direction for π/K, here takes K=8. to express the fingerprint image by each picture element direction, as shown in Figure 2.
Then and type (2) extracts 8 directions separately according to type (1) the Sd value.
And f(i, j) is (i, j) grey level, (ik, jk) is on the direction d kth spot, N is the direction number which takes, n is the neighbouring points which takes, here takes n=4, d’daibiao and the d vertical direction. (i, j) direction is S4 value smallest direction d, namely the gradation changes the smallest direction. Then, supposes N(d) is in some picture element 8 neighborhoods the direction is the d picture element integer, defines N(d) the maximum value is N(D1), a great value is N(D2), its corresponding direction respectively is D1 and D2, C (i, after j) is (i, i) adjustment direction code, carries on smoothly according to type (3), thus obtains the good filter effect.
Then carries on binaryzation processing, binaryzation processing is refers to view picture imagery processing Cheng Jincun black (grey level is 0) and white (grey level is 1) two values. This article uses the dynamic valve value law, divides into certain subblocks the fingerprint image to calculate each subblock the gradation average value, if this in some picture element grey level is bigger than this grey level, then this spot takes 1, otherwise takes 0, carries on processing according to this law to each subblock. In binaryzation process unavoidable introduction noise, therefore, after the processing, must carry on takes chirp processing, the value which after its function is fills binaryzation, the grain on-line hole or the deletion grain on-line burr and encourages is 1 picture element.
Then the opposite becomes two value images carry on refinement processing, namely, in does not affect in the image connective foundation to delete the grain line the edge picture element, until texture line for single picture element. The author selects the cascade iterative algorithm, and makes an iterative minute two this scanning, by peripheral to middle cascade refinement, causes the refinement result to be located at the original map the middle axle, then must carry on denoising processing.
Figure 3 shows fingerprint image after pretreatment effect which the fingerprint gathers.
3.2 extraction characteristic points
The fingerprint recognition algorithm summed up finally for finds and contrasts the fingerprint in the fingerprint image especially on time. Discovers all connection number after the pretreatment fingerprint image is 3 spots, may obtain the bifurcation; But seeks the vertex to use the method which one kind is accommodating the fingerprint division chart opposition, the refinement, then withdraws the bifurcation again, obtains the bifurcation is the actual vertex. Like this obtains the characteristic point majority is really, but also contains the false characteristic point which produces because of the noise effect, this will cause the appraisal probability of misrecognition rise, must therefore remove these false characteristic points. Edge some characteristic points and some cavities and burr elimination: The computation already designated possible characteristic point and edge distance, when is smaller than threshold value time, thought that this spot is unreliable, its elimination. To the central point segment and the fingerprint image coordinate’s included angle attribute the positional information with the characteristic point and between the fingerprint central point’s grain line number as well as the characteristic point, stores the fingerprint storehouse these positional information.
3.3 fingerprint matches
Between the localization input characteristic set of points and the existing pattern plate characteristic set of points’s match zero point is right, calculates two characteristic sets of points revolving and the translation parameter. And acts according to these revolving and the translation parameter calibration input set of points, after will calibrate the characteristic set of points transforms to the limit in the coordinate system, then between comparison characteristic point reciprocity set, obtains between the fingerprint image and template’s corresponding relationships.
4 hardware architectures
System hardware structure as shown in Figure 4, the system core selects. I’I Corporation’s TMS320VC5409 the DSP chip, it is 16 bit fixed-point DSP, has the good performance-to-price ratio, its performance may satisfy this fingerprint appraisal system request. Its main feature is as follows:
* the operating speed is quick, reaches 100 MI/s;
* optimized CPU structure;
* the intelligent peripheral device, the expansion is convenient;
* low power loss.
DSP from the fingerprint sensor read fingerprint image data and carries on corresponding algorithm processing, will then process the result to deliver to the liquid crystal display module demonstration. Using the DSP parallel memory connection, the multichannel cushion serial port takes and the fingerprint sensor, the LCD display monitor and the pressed key and so on connection. In order to satisfy the system request, has disposed SRAM and EEPROM for DSP, and selects CPLD to complete overall system’s logical design.
The author was DSP expanded 64 KB CY7C1021 SRAM, but DSP most supported exterior expands 32 KB data space, needed to control SRAM through the CPLD in logic circuit a15 address wire level, thus decided that was chooses SRAM in the same time high address section 32 KB the storage space low address section 32 KB storage space. Disposes SST29EEOlO EEPROM satisfies DSP BOOTLOADER the need, on after the system the electricity, DSP first to carry out the BOOTLOADER procedure. In the EEPROM procedure loads DSP outside to DSP internal RAM in the high speed execution, like this, the system may the completely off-line working, not only may reduce application system’s hardware cost, but may also enhance system’s overall resistance to interference.
This system selects American ALTERA Corporation’s MAX7000 CPLD, this is the present speed quickest high integration rate programmable logic component series. The author has designed logic which using CPLD the many system coordinated work need, has used many DSP condition attribute signal participation logical design, moreover may by cooperate the simple treatment part the processor to complete.
The fingerprint gathering installment uses the FPS200 electric circuit which American VERIDICOM Corporation produces, it is a contact high performance, the low power loss, the low price electric capacity type fingerprint sensor, the principal characteristic is as follows:
* uses the standard CMOS craft manufacture, 500 d/i high resolution.
* seal size: 24 mm*24 mmx2.5 mm;
* sensing region: 1.28 cm*1.50 cm;
* sensor array: 256×300;
* the surface has the ultra hard coating protection;
* built-in 8 A/D switch. But direct output digitization image signal. The power loss does not surpass 200 mW;
* has the automatic fingerprint measuring ability.
The LCD module uses FMl28641 to take the fingerprint graph and the processing result demonstration user interface. FMl28641 is one kind of graph lattice liquid-crystal display. It is mainly composed of a good driver/row driver and the 128×64 entire lattice liquid-crystal display, may complete graphical display, may also demonstrate 8×4 (the 16×16 lattice) the Chinese character. For ease of the debugging, the company enters first in the circuit design the DSP output and the LCD connection’s control pin CPLD, is advantageous adjusts the succession nimbly.
5 experimental results and debugging
This fingerprint recognition system’s outward appearance is exquisite, the operating performance is stable. The algorithm development earlier period uses the C language to carry on the functionality confirmation for the algorithm. After confirming successfully. The use carries on the algorithm development face the TMS320VC5409 DSP assembly language in the CCS environment.
The fingerprint image’s pretreatment and the match procedure use the software simulation, the simulation environment are SIMULATOR first. In the algorithm correct situation, carries on the hardware simulation again, gathers the fingerprint image to carry on the debugging directly. Hardware’s each module programmed the debugging, the environment is also alone SIMULATOR. has obtained the good timeliness.
