We present a study on pulse shapes of a CdZnTe semiconductor detector for beta particles and gamma-rays emitted by a ¹³⁷Cs-⁹⁰Sr mixed source. When beta and gamma radiations enter a semiconductor detector, two kinds of pulses are produced according to the type of radiations. The characteristics of the pulses are determined by their energy and depth of interaction. The energy deposited on the detector basically determines the pulse height and the depth of interaction affects the pulse shape. Because of the different characteristics of beta and gamma radiations, their pulse shapes are also different. The digital-analog hybrid signal processing method was applied to separation of the beta spectrum from the beta-gamma mixed one. The separation of beta spectrum was done based on a neural network algorithm, a leading edge area method, and a falling continuum tail value method. However, the perfect separation could not be obtained. The separation was also done by reconstructing the spectra on the basis of subtraction of gamma-ray component. The spectra of beta particles and gamma-rays from ¹³⁷Cs-⁹⁰Sr mixed source could be separated well by using the reconstruction method.