An on-line bearing fault diagnosis technique based on passive vibration sensor tag is proposed.Structure of a novel passive radio frequency identification（ RFID） vibration sensor tag is designed

which can realize longt time and on-line fault diagnosis and decrease cost due to its passive operation mode.Processing mode of vibration signal is introduced. Singular value decomposition（ SVD） signal denoising algorithm based on singular entropy is proposed

according to singular entropy of signal to determine order by itself and denoise

avoid error caused by man-made preset parameters

and fault diagnosis algorithm based on least squares support vector regression（ LSSVR） is proposed.Test results show that the designed tag can achieve signal acquision and transmission reliably

the employed algorithm can locate fault fastly and accurately

which improves the real-time performance and decrease the cost compared with traditional fault diagnosis method.