The invention discloses a construction method for a cross-equipment electromagnetic
fingerprint database on the basis of
machine learning. The method comprises the following steps that: firstly, collecting the
electromagnetic signal of equipment, and carrying out denoising, specific window bandwidth interception and digital
processing on the
electromagnetic signal; then, carrying out the
machine learning on the processed
electromagnetic signal, judging and obtaining the classification and the marking of the electromagnetic
signal, and conveying the classification and marking information to a display; and finally, judging the
correctness of a displayed equipment identification result by a user, if the displayed equipment identification result is correct, storing the classification and marking information into the electromagnetic
fingerprint database, otherwise, collecting the electromagnetic
signal of the equipment again, and carrying out
machine learning again until the displayed equipment identification result is correct. By use of the method, the electromagnetic
signal of the equipment can be accurately obtained so as to establish the electromagnetic
fingerprint database, and the electromagnetic
fingerprint database can be updated and enlarged while the database is used. The invention also discloses a device for realizing the above method, the support of external marking equipment is not required, cost is saved, and
natural interaction with articles is enhanced.