Abstract: A novel method based on concurrent time-frequency atom（TFA） complex band-pass filters was proposed in this paper for online monitoring low frequency oscillation.First several TFA complex band-pass filter units were arranged within the low-frequency band to obtain mode magnitude,frequency and attenuation factor.Second,particle swarm optimization（PSO） algorithm was utilized to estimate the phases and DC component.Each filter has flexible adjustability of its time window and frequency bandwidth.Therefore,high mode resolution and short observation time window can be self-adaptively achieved.The proposed method can identify composite modes and strongly time-varying modes accurately with good noise robustness,catering for online analysis.The simulation results and the field test result of a real power system all show its correctness and effectiveness.