Abstract: The current methods for analyzing the nonstationary and nongaussian TWA of automatically detecting and estimating T wave alternans (TWA) in multilead ECG have poor sensitivity to nonstationary and lowamplitude. The particle filtering (PF) method combined with joint analysis information was proposed to detect and estimate multilead TWA. According to a multilead TWA analysis scheme， a linear transformation using principal component analysis was processed on the preprocessed multilead STT complex. Then PFbased method was applied to each lead for TWA detection and estimation. To evaluate the proposed method， multilead PF method was compared to single lead PF and spectral method. The simulation results showed that the proposed method had the best performance on detection and estimation of TWA when the alternanssignaltonoise ratio was lower than -15dB. For nonstationary and nongaussian TWA， the proposed method had good robustness and high antinoise ability.

Key words: multilead ECG, T wave alternans, particle filtering, PCA, Laplacian noise