Detecting the marine phytoplankton by the means of absorption or fluorescence spectra were successfully deployed in the past decades, however, the differentiation are mainly limited in levels of class, such as bacillariophytas, dinophytas, raphidophytes, chlorophytes, cyanobacteria, etc. which are characterized by their specific composition of photosynthetic pigments. To further differentiate the typical dinoflagellate Prorocentrum donghaiense, Amphidinium carterae, Scrippsiella trochoidea, Karenia mikimotoi out of the common diatom Skeletonema costatum and haptonema Phaeocystis globosa at East China Sea, a rapid 3D-fluorescence method equipped with CHEMTAX model were conducted. Initial fluorescence excitation spectra of each algal species (under variable environmental conditions) were captured by 3D-fluorometer first. Then fingerprints of each algae were characterized by ten-point discrete excitation spectrum with the excitation wavelengths of 405, 420, 435, 470, 490, 505, 535, 555, 570 and 590 nm, which closely reflecting the difference of photosynthetic pigments. By equipping with CHEMTAX model, the standard spectra and norm spectra were constructed for FS-CHEMTAX (Fluorescence spectra-CHEMTAX) model to further identify the algal species and estimate the cell density. The developed method performed a better way of identifying the toxic species Amphidinium carterae, Phaeocystis globosa, and Karenia mikimotoi out of the non-toxic ones, with the identification accuracy rates of 83.3%, 90% and 100%, in monocultures, and 77.8%, 90% and 100%, in the bi-mixed cultures, respectively. Meanwhile, the detection limits for the three toxic species were found as low as 250, 1,400 and 120 cells/mL. The concentrations estimated are in good agreement with the microscopic cell counts for all the algae groups (correlation coefficients (R²) exceed 0.8). The relative error of predict concentration was lowest for small cells, i.e., Phaeocystis globosa (10.0%) and Amphidinium carterae (21.1%), but the highest for big cells, i.e. Karenia mikimotoi (41.8%) when the target algae become the dominant species. The overall concentration detection error was no more than one order of magnitude, indicating that this method could provide an important technical support for monitoring the related harmful algal blooms.