The phenomenon of frequency sensitivity in weak signal detection has been studied via numerical simulation in the cases of a single neuron and a neuronal network, respectively. The membrane potential undergoes a damping oscillation to the quiescent state with a frequency of 11--33 Hz under a subthreshold bias. In the presence of a subthreshold sinusoidal signal and noise, the signal-to-noise ratio (SNR) of either the single neuron or the network presents the typical characteristic of stochastic resonance. In particular, there exists a frequency sensitivity range of 15--60 Hz, over which the SNR has a large value, implying that the system is more sensitive to signals with these frequencies in signal detection. Such frequency sensitivity is of functional significance, and results from the resonance between the intrinsic oscillation of the system and the input signal. In addition, two cases where either the values of the bias or the signal strength for neurons may be different are studied, respectively. The results indicate that the effect of frequency sensitivity generally exists in signal encoding. These enable us to interpret the experimental observation.