The data reductions of existing borehole seismic methods are currently based on travel-time analysis in which first-arrival times of the seismogram are manually picked. However, manual picking is not efficient and depends heavily on the data quality and analyst's experience. Besides the difficulty of picking first arrivals in the time domain, no information is obtained regarding the effective frequency of the analysis. Data reduction method based on the time-frequency domain was proposed in the study. The aim of the study was to propose a semi-automatic procedure for more objective first-arrival time picking. Evaluation using field data shows that the semi-automatic time-frequency spectrum analysis works well in PS-Log testing. The picked first arrival times from the proposed method agree well with the manually-picked results. The procedure is stable and yields reasonable results. But it should be noted that the method inherit the tradeoff between frequency and time resolution from the time-frequency spectrum analysis. The parameters used in the time-frequency analysis should be optimized for different source types.