This paper proposed the signal processing method for evaluation of arteriovenous shunt (AVS) stenosis in hemodialysis patients. AV shunts are surgically created pathological-physiological fistulas serving as access routes for end-stage renal disease (ESRD) patients. The distinct and periodic bruit of vascular shunts is clearly audible over the access routes. Thus, a bruit spectral analysis could be considered a valuable noninvasive method for quantifying the severity of vessel stenosis. This study collected phonoangiographic data from thirty AV shunts obtained from the electronic stethoscope during the pre- and post-percutaneous transluminal angioplasty (PTA) periods. An autoregressive (AR) model was applied to scientifically analyze the phonoangiographic signals. The AR model and the filter order of eight were chosen to estimate the characteristic frequency of the bruit. The AR model results obtained from the analysis of the phonoangiographic data under the pre-PTA and post-PTA conditions show significant changes in frequency and magnitude. These findings could serve as a noninvasive strategy for early detection of stenotic lesions.