Application of ARM core AT75C220 in fingerprint recognition system

In this way, AT75C220 is very suitable for devices that require image processing functions and have Internet communication capabilities, such as high-performance fingerprint recognition devices and VoIP applications. The integration of multiple functions on a single chip can reduce costs and improve system reliability. At the same time, the chip's complete development software support (including LINUX operating system and DSP algorithm program support, etc.) is also conducive to reducing product time to market.

2 Chip Structure

AT75C220 is mainly composed of RISC processor based on ARM7TDMI with running speed of 40 MI/s, DSP processor with running speed of 60 MI/s and dual Ethernet 10/100 Mbps MAC interface. Its internal structure is shown in Figure 1.

The ARMTTDMI microprocessor is a high-performance embedded CPU with low power consumption and fast computing speed. It contains a Thumb instruction decoder that supports 16b instructions, debugging and fast multiplication, and can exchange data with other functional modules through the module intercommunication bus. ARMTTDMI
(Thumb) is the product with the most authorized users of the company. It combines the ARM7 instruction set with the Thumb extension to reduce memory capacity and system cost. At the same time, it also uses embedded ICE debugging technology to simplify system design and uses a DSP enhancement extension to improve performance. Another reason for choosing products based on the ARM core is that ARM's processor core has been supported by a large number of manufacturers, which is convenient for product development and upgrading.

Compared with other chips with integrated DSP cores, the DSP subsystem resources of AT75C220 are quite rich. As shown in Figure 2, in addition to a 60MI/s, 16b fixed-point DSP core, it also has RAMs for different purposes: 2kB×16X-RAM, 2kB×16Y-RAM, 16kBX16 general data RAM, 24 kB×16 downloadable program RAM, and a 256×16 dual-port RAM mailbox for MCU-DSP communication. The complete CODEC interface allows a large number of industrial voice, multimedia or data connections. [page]

As a branch of pattern recognition research with relatively high difficulty, automatic fingerprint recognition mainly focuses on how to filter and enhance various noisy images collected, how to extract global and local features of fingerprints, and how to match features when the image cannot be completely located or may be distorted. Generally, it can be divided into the following steps. As shown in Figure 3, the fingerprint recognition algorithm process mainly includes grayscale filtering, binarization, binary filtering, refinement, denoising after refinement, and finally feature extraction and matching.

4 Fingerprint recognition system

The fingerprint recognition system is generally composed of a fingerprint sensor, a DSP processor and peripheral circuits. In order to increase the processing power of the system, the system often includes other processors to manage such as communication, human-machine interface, fingerprint database, etc. The structure of this system is shown in Figure 4. The capacitive fingerprint sensor FPS110 collects fingerprint images and sends them to the DSP coprocessor inside the AT75C220 for fingerprint recognition, completes image preprocessing, extracts fingerprint feature codes, and matches them with fingerprint features in the fingerprint library to give a judgment result. The ARM core runs on the LINUX operating system and implements the TCP/IP protocol stack. Since the network system of the fingerprint identifier does not require high real-time and throughput, but requires high reliability and confidentiality, the transport layer protocol selects the TCP protocol, and the transmitted data is encrypted. The ARM processor also includes a human-machine interface and control output, which is used to control the electronic door lock. In addition, the ARM software also includes some proprietary control modules in the access control system and attendance system, such as access control authority management and attendance database management.

5 Algorithm transplantation

The fingerprint recognition algorithm was initially implemented on a PC and needed to be ported to the DSP subsystem of the AT75C220. If it was ported directly in C language, it would be difficult to give full play to the actual processing power of the DSP and the execution efficiency would be quite low. After the key algorithm modules were implemented in assembly language, fingerprint collection, image preprocessing and matching could be completed in 1 second. The algorithm efficiency and recognition rate were comparable to those calculated on a PC, meeting the requirements of actual applications.

6 Conclusion

Since the hardware is composed of the integrated multi-functional AT75C220, the system hardware design is simple, making the entire system meet the requirements of miniaturization, networking, and intelligence. Practice has proved that the processing capacity of equipment using this structure has been improved, representing the future development direction of access control systems.

The integration of hardware functions makes the main design work focus on software design. How to further improve the recognition rate, improve the distributed database system, and ensure the confidentiality of data transmission are the key points of future design.

References:

[1]. RS485 datasheet http://www.dzsc.com/datasheet/RS485_585289.html.
[2]. AT75C220 datasheet http://www.dzsc.com/datasheet/AT75C220_810052.html.
[3]. ARM7TDMI datasheet http://www.dzsc.com/datasheet/ARM7TDMI_139812.html.
[4]. 16b datasheet http://www.dzsc.com/datasheet/16b_2177774.html.
[5]. RISC datasheet http://www. dzsc.com/datasheet/RISC_1189725.html.