Discrete wavelet transform has been proposed for the diagnosis of paravalvular leakage(PVL) occurs after Mechanical Heart Valve(MHV) replacement. Mechanical heart sounds of 2 patients with PVL and 5 volunteers with normal MHV was recorded and analyzed with different wavelets. As a result, it is statically shown that Daubechies 2 wavelet function is suitable for diagnosis of PVL.Paravavular leakage (PL) is a situation when part of the mechanical valve ring remains disconnected from the recipient's annulus. Mechanical heart valve leakage is a relatively rare complication in patients undergoing heart valve replacement, but patients with severe paravalvular leakage usually have symptoms of heart failure or severe anemia and should be treated with surgical repair or replacement of the valve [1]. In this study, relationship between features obtained from discrete wavelet transform (DWT) of mechanical heart valve (MHV) sound signals and paravavular leakage was investigated.For this aim, MHV sound signals of two patients with mitral paravalvular leakage (MPL) and five patients with normally functioning MHV. All volunteers had mitral heart valve replacement (MVR) with bileaflet MHV. The heart sounds of two patients with leakage were recorded before and after repair of MPL. The electrocardiogram signal (ECG) and heart sound signals of patients were simultaneously recorded. Detailed information of recording system and procedure can be found in [2]. Recorded ECG and heart sound signals were normalized and filtered with a 30 Hz high pass and 2000 Hz low pass filter. After that the heart cycles of MHV sounds were detected using ECG signals [2]. As a result of detection procedure, MHV sounds of 147 heart cycles of patients with MPL, MHV sounds of 96 heart cycles of same patients after treatment of MPL and MHV sounds of 98 heart cycle of patients with normally functioning MHV were obtained.Discrete wavelet transform (DWT) of MHV sound signals of all detected heart cycles were computed using the MATLAB software tool. The DWT of the MHV sound signals are computed by using five different wavelet functions: Daubechies 2, Daubechies 5, Daubechies 10, Symlet 5, and Symlet 7. The number of decomposition levels was chosen as 5. Thus, the heart sound signals were decomposed into the details D1-D5 and one final approximation, A5. The frequency bands corresponding to different levels of decomposition for a sampling frequency of 5000 Hz are 1250-2500 Hz for D1, 625-1250 Hz for D2, 312.5-625 Hz for D3, 156.25-312.5 for D4, 78.125-156.25 Hz for D5 and 0-78.125 Hz for A5.Sub-band signals obtained from DWT of one heart cycle was split into four parts relating to physiological properties of cardiac cycles. Each cardiac cycle is composed of two phase: contraction (systole) and relaxation (diastole). So heart cycle can be split into four parts using physiological properties like that: PartS1 represents S1, PartS2 represents S2, PartSis represent systole area except S1 and PartDias represent diastole area except S2. Table I. MEAN AND STANDART DEVIA...