With the popularity of fingerprint recognition in smart phones, fingerprint recognition technology has been applied in more and more occasions. In addition to mobile applications, mobile payment, access control system, smart home and other embedded scenarios are also gradually popularized. In the system implementation, smart phone itself has a strong computing power and rich memory resources, fingerprint recognition is not difficult to achieve, but in embedded system especially based on MCU application, its computing power, memory resources are limited

 

As shown in the figure, this is a system block diagram of the front end of WiFi fingerprint recognition. Its application scenario is as follows: when the finger presses the fingerprint recognition module, the fingerprint data is collected and transmitted to the MCU, which processes the fingerprint data through the identification algorithm, and then wirelessly transmits the processing results to the cloud through the WiFi module to participate in the identity identification business.

                                                                                    

 

In terms of specific implementation, the fingerprint recognition algorithm involves a lot of floating point arithmetic and needs to temporarily store the original data and intermediate data of the fingerprint lattice, so it has rigid requirements on computing power and storage space. In the current mainstream single chip microcomputer architecture, the Cortex-M4 architecture integrates FPU floating point processing unit, and at the main frequency of 100MHZ, Its floating-point arithmetic ability can meet the requirements. The fingerprint recognition algorithm code takes up hundreds of bytes of code space after compiling. Considering the WiFi network connection, application layer code and other aspects, the overall appropriate Flash code space is 1MB, and the data storage requirements are 512KB of SRAM space. When the system works, it needs strong computing power in the process of fingerprint recognition, but when there is no fingerprint press, it needs to run in the state of low power consumption, so as to adapt to the requirements of embedded system on power consumption.

 

In our scheme, we choose an MCU with XIP characteristics, store the code in external SPI Flash and execute it in the system, thus greatly expanding the code storage space. The execution speed of the external SPI Flash is slightly lower because the code in the SPI Flash requires an internal Cache. The code that identifies the algorithm can be copied to the SRAM during the Boot phase to improve the running speed. The XIP SRAM code space allocation scheme, which addresses both performance and cost, is a highlight of this design.

 

The fingerprint identification chip is the component of the system implementation. The current mainstream technical indicators require the fingerprint identification chip to be based on capacitor technology, support Live Finger DetecTIon, press type, adopt glass cover plate, realize 360 degree touch in any direction, and support sliding navigation. In vivo detection technology can prevent fake fingerprint cracking and integrate heart rate detection function, which is convenient for users to check the heart rate value in real time. Sensor integration is convenient for the design and integration of the whole machine manufacturer, which makes it easy to be applied in the embedded single-chip microcomputer system.

 

The data interface between the fingerprint identification chip and the host requires a transmission rate of more than 5Mbps during fingerprint data collection. If the data throughput is lower than 5Mbps, the user experience will be affected. The low-speed interfaces such as UART and I2C cannot meet the requirements in terms of throughput, while the SPI interface is simple and the transmission rate can fully meet the requirements, so it is a suitable communication interface.

 

The fingerprint recognition chip periodically detects whether there is finger touch in the Pixel sensor area. When finger touch is detected, the live detection data will be collected immediately. After data collection is completed, the Host will be informed to read through interruption. When the fingerprint scan begins, the Host is notified to read the data through an interrupt. Fingerprint data is scanned and Host data is read simultaneously.

 

After receiving the original Pixel data transmitted by the fingerprint chip, the MCU calculates it through the identification algorithm and transmits it to the cloud through the WiFi module. Due to the limitation of MCU's own resources, WiFi module itself needs to integrate WiFi driver and TCP/IP protocol stack, and can be used as a relatively independent unit to run application layer code, which greatly reduces the burden on MCU host side. The WiFi module interacts with MCU through serial port.

                                                                                    

 

Low power consumption and anti-interference are also a key point of system design. The fingerprint recognition module still performs periodic sensor scanning when there is no finger pressing. Although the power consumption is lower than that during data transmission, it is hoped that the power supply of the fingerprint recognition module can be turned off when there is no key touch in order to adapt to the occasions of battery power supply. When a finger is detected near, turn on the power supply of the fingerprint identification chip for fingerprint scanning and collecting data. When no finger touches for a long time, turn off the power supply of the fingerprint scanning module to reduce the power consumption. In addition, in order to prevent the impact of touching the key on the fingerprint recognition sensor during data collection, an analog switch is added after touching the key, and MCU outputs a control signal to isolate the analog signal of touching the key before the collection begins.