RANDOM NOISE CANCELLATION IN BIOMEDICAL SIGNALS USING VARIABLE STEP SIZE GRIFFITH LMS ADAPTIVE LINE ENHANCER

This paper presents a new method of random noise cancellation for removing artefacts from biomedical signals using an adaptive line enhancer (ALE). The ALE is implemented using proposed time domain variable step size Griffith least mean square (VSGLMS) algorithm. The technique is based on the adaptation of the gradient of the error surface. The method makes both the step size and the gradient free from observation noise and reduces the gradient mis-adjustment error. Here, both the gradient and the scale factor for the step size are free from the input noise effects, which makes the algorithm robust to both stationary and non-stationary observation noise. Further the additional computational load involved for this is marginal. The VSGLMS adaptive filter technique for ALE is tested on noise cancellation of two types of biomedical signals - separation of electro cardiogram (ECG) signal from a background of electro myogram (EMG) and heart sound signal (HSS) from lung sound signal (LSS). Application of VSGLAM - ALE for the separation of HSS from LSS and ECG from EMG has been demonstrated using synthetic White Gaussian noise (WGN). It is found that VSGLMS - ALE can separate the desired signals like ECG or HSS at an input SNR of -5 dB to 27 dB. The performance of VSGLMS is compared with state-of-the-art least mean square LMS - ALE and normalised LMS - ALE. The results of PSDs, time domain waveforms, and mean square error (MSE) have proven that VSGLMS performs better than advanced techniques.