The fingerprint module selection is not a problem of “looking at the parameter table”, the core is to set the scene first, then look at the indicators, and finally check the samples.

 Step 1: Determine the application scenario first (decide the technical route) Also called “fingerprint module”, the selection logic of different scenarios is completely different:

Application scenarios	Recommendations for selecting FP sensor	Select the key points
Smart door lock/cabinet lock	Medium and small array semiconductor capacitive sensor	Anti-fake fingerprint, small size, living body detection, low power consumption, the larger the chip matrix, the better the recognition, considering the balance of cost and array size, 160 * 160 below can be, but too small array (for example, 80 * 64 below) is not recommended for poor experience.
Access Control /Attendance (High frequency be used)	Optical type	Wear resistance, long life and low cost, but the optical scheme can be deceived by a duplicate silicone fingerprint, and the security is low.
Financial POS/safe deposit box (high security)/medical equipment/public security or bank authentication	Capacitive /large array sensor	Very low FAR, anti-fake fingerprint, small size, live detection, low power consumption, such as 208 * 288, 256 * 360 matrix etc.
Portable/Battery Powered	Capacitive sensor	Ultra-low power consumption and small size
Smart phone	Capacitive/ultrasonic/ fingerprint under screen	Oil and wet hand recognition, ultra-thin, ultra-low power consumption, small size

Step 2: Understand the core indicators (don’t just focus on the nominal value of FAR/FRR)

Specification	Meaning	Reference value	Remark
FAR	Probability of mistaking different fingers for the same finger	Door Lock < 1/50000 (0.002%), Finance < 0.001%	Look at the test conditions, not just the nominal
FRR	Probability of my finger being rejected	Door Lock < 1/100 (1%), Finance < 1/100 0(0.1%),	Performance under dry/wet/dirty hand conditions is more important
Sensing area/pixel	Look at product demand	FAR and FRR of small area are worse, while that of medium and large area is much better, but the cost is directly related to the area.	The area registered for more than 3 times is small area, which is mainly based on image algorithm, and the area registered for 3 times is medium and large area (more than 160 * 160), which is mainly based on eigenvalue algorithm.
Compare the speed	It takes time to identify once	< 0.5s is preferred, the sooner the better	Directly affect the experience
Fingerprint capacity	How many pieces can be saved?	There are 10-20 cabinet locks, 50-100 door locks, more than 1000 medical equipment/public security or bank identity authentication, and up to 10000 local storage.	Don’t blindly pursue large capacity, which directly affects the cost and identification speed.
Service life	Number of compressions	Capacitive about 200 thousand times, optical about 1 million times	High frequency gate disabling optics is more cost effective
Power consumption	Operating/standby current	Battery products focus on machine power consumption	Normally, it needs to last for more than 12 months.

Step 3: Confirm the interface and power supply (corresponding to your electronic board design)

Interface	Applicable scenarios	Explain
UART	The vast majority of locker solutions	Most common, serial communication, least wiring
USB	Attendance/Financial Equipment	Drive-free, direct connection to PC

Step 4: Proofing Verification List

1. Identification experience: dry hands, wet hands, sweaty hands, dirty hands and fingers with shallow wrinkles of the elderly were measured 50 times respectively, and FRR was counted.
2. Anti-counterfeiting test: Use the printed fingerprint/silicone fake finger to test, and the capacitance scheme shall be rejected.
3. Temperature and humidity: run once in high and low temperature and humid environment (the capacitor scheme has higher requirements for the environment and is afraid of dust and moisture)
4. ESD/static electricity: The lock body is often in contact with the human body, and the static electricity will be hit to see if it will crash.

Actual measurement of standby power consumption: confirm that the battery life is up to standard 6) Consistency: Take 3-5 samples to cross-check whether the recognition rate is stable.