In cognitive radio system, it is necessary for spectrum sensing technology to search for vacant channel. Cyclostationary feature detection method has superior noise-immune and recognition. However, since the mismatch between the detected signal and detection duration by guard interval (GI) insertion occurs, the cyclostationarity is degraded. In this paper, we propose a high accuracy and low complexity detection method based on cyclostationarity feature. The proposed method limits symbol detection time and makes multiple delay line detectors as units of the GI time. Computer simulation and analytical evaluation show the effectiveness of the proposed detection method compared to the conventional one.