ECG Denoising Using Cubature Quadrature Kalman Filter Approach

Noise contaminated electrocardiogram (ECG) denoising is the crucial pre-processing stage in the analysis of cardiac diseases, as analysis and diagnosis get affected by these contaminations. This paper presents a proficient and robust recursive filtering technique under the Bayesian framework based on the cubature quadrature Kalman filter (CQKF) is proposed for ECG denoising. The proposed framework utilizes angular velocity as the third-state in the dynamic model, and methodology is termed as CQKF3. The performance of CQKF3 is estimated on five ECG records (with no significant arrhythmia) randomly chosen from the MIT -BIH normal sinus rhythm database. The efficacy of the CQKF3 framework is tested on non-Gaussian (real muscle artifacts) and Gaussian noise environments in the range of input -5 to 10dB signal-to-noise ratio (SNR). The CQKF3 is compared with extended Kalman filter (EKF) frameworks that are EKF2 and EKF3, and obtained results disclose that the proposed framework gives a better improvement in SNR under both unwanted contaminations.