In this paper an efficient spectrum sensing technique for burst signals which appear at a specific channel frequency raster has been proposed. It is based on the cross-correlation of intra-segment decimated vectors and the channel occupancy decision is based on FFT of the correlated output. The proposed technique is different in that sensing decision is based on FFT output bin with maximum magnitude. This increases the probability of detection (Pd) and reduces the probability of false alarm (Pfa) for burst signals. Computer simulations with multiple narrowband QPSK modulated signals have been carried out to validate the proposed technique's performance. Performance in comparison to FFT only approach is done in AWGN, Rician and Rayleigh fading multipath (6-path) channel environments. The proposed technique is able to achieve a Pd of over 90 percent with Pfa of less than 5 percent for sensing a burst of 10msec duration with a sensing time of 3.6msec at 0dB SNR for all the simulated environments. The proposed technique's detection performance is found to be similar when multiple bursts at different raster frequencies are simultaneously injected. Further, an efficient implementation architecture for the proposed spectrum sensing technique with multiple instances of the architecture for parallel sensing of multiple bursts has been presented.