Ultrasonic techniques are widely employed in the nondestructive characterization of materials. For example, the use of grain backscattered ultrasonic signals for the estimation of grain size has been studied extensively [1,2,3,4]. Several techniques to process the grain backscattered signals and extract information related to grain size have been reported in [4]. In this paper, we describe a new technique to process these signals and extract features that can be used for material characterization. The technique consists of the following three steps: i) deconvolution of the backscattered signal to remove the effect of the measurement system, ii) estimation of the spectrum of the resulting reflection coefficient sequence, and iii) extraction of features from the spectrum related to the average scattered energy and the rate of change of scattered energy with frequency, both computed within the bandwidth of the ultrasonic transducer. The spectral features so extracted are influenced by the microstructural properties of a material pertaining to scattering, e.g., average grain diameter, and can be used in the characterization of these properties.