A clean signal is an ultimate prerequisite to obtain good results and correct understanding of a physiological process. Noisy signals can lead to false diagnosis or misinterpretations. ECG signal is sensitive in nature and gets affected by different noise types during acquisition. This paper presents a hybrid denoising algorithm based on Empirical Mode Decomposition (EMD) and Wavelet Transform (WT) for denoising electrocardiogram (ECG) signals. The algorithm performs Empirical mode decomposition and decomposes the signal into intrinsic mode functions (IMFs). Selected noisy IMF is decomposed by wavelet transform and IMF coefficients are thresholded to remove the noisy components. Denoised IMF is reconstructed by taking inverse wavelet transform and added back to the signal. The algorithm is tested over stress ECG dataset corrupted with baseline wandering and electrode movement artifacts. The algorithm is further tested by adding synthetic white Gaussian noise to the signal in the range 5-20 dB. Signal-to-noise ratio (SNR) and Covariance parameters are used to evaluate the performance of the proposed algorithm with the existing standard methods. The subjective and objective comparison suggests a better performance by the proposed algorithm in comparison to its standard counterparts.