[Discussion about the prediction accuracy for dynamic spectrum by partial FFT].

The development of near-infrared-based techniques for the noninvasive determination of blood component concentrations has attracted significant interest in recent years. But the noninvasive measurement of blood compositions has not yet been applied to the clinical field except blood oxygen saturation. The most important and also difficult problems are the effects of individual discrepancy and complicated measurement conditions. In the present article, the approach of dynamic spectrum (DS) is introduced, which is based on the principle of photoplethysmography. It is very difficult too to pick up DS with high precision in time domain. In order to extract the DS with high accuracy, the FFT method and it's leakage are discussed. The influences of sampling speed, sampling signal periodicity, window function and unsynchronized sample are analyzed by emulating experiments. The result of emulating experiments shows that choosing certain sampling speed, sampling signal periodicity, window function and interpolation arithmetic will improve accuracy observably. This provides necessary condition for the clinical application of DS.