Abstract We introduce a new shrinkage scheme, hyper-trim that generalizes hard and soft shrinkage proposed by Donoho and Johnstone (1994). The new adaptive denoising method presented is based on Stein's unbiased risk estimation (SURE) and on a new class of shrinkage function. The proposed new class of shrinkage function has continuous derivative. The shrinkage function is simulated and tested with ECG signals added with standard Gaussian noise using MATLAB. This method gives better mean square error (MSE) performance over conventional wavelet shrinkage methodologies.