Audio fingerprinting systems are often well designed to cope with a range of broadband noise types however they cope less well when presented with additive noise containing sinusoidal components. This is largely due to the fact that in a short-time signal representa- tion (over periods of ≈ 20ms) these noise components are largely indistinguishable from salient components of the desirable signal that is to be fingerprinted. In this paper a front-end sinusoidal noise reduction procedure is introduced that is able to remove the most detrimental of the sinusoidal noise components thereby improving the audio fingerprinting system’s performance. This is achievable by grouping short-time sinusoidal components into pitch contours via magnitude, frequency and phase characteristics, and identifying noisy contours as those with characteristics that are outliers in the distribution of all pitch contours in the signal. With this paper’s contribution, the recognition rate in an industrial scale fingerprinting system is increased by up to 8.4%.