In this paper, we propose a detection algorithm to detect unknown multitone interference (MTI) encountered in frequency-hopped spread spectrum (FH-SS) systems over a Rayleigh fading channel and additive white Gaussian noise by using autoregressive (AR) spectral estimation method. The proposed algorithm is performed in the AR coefficient domain. The incoming received data is separated into non-overlapping data segments, and the Yule-Walker (YW) AR method is applied on each data segment to obtain the AR coefficient vector. The resultant coefficient vectors are used in the averaging process to obtain the spectrum estimate of the MTI. Computer simulations are conducted in order to investigate detection performance of the proposed AR spectral averaging technique and comparisons are made between the proposed scheme and the fast Fourier transform (FFT) based scheme. Numerical results show that the proposed technique is able to detect the MTI accurately under such a channel environment even at low signal-to-noise power ratio (SNR) condition. The comparisons reveal that the proposed AR detection technique is promising to be employed in slow FH-SS systems to detect the MTI due to higher estimation accuracy and better reliability as compared to the FFT- detection technique.