Transmission of information from within the body to the external environment has become a popular issue in making the Internet of Nanothings (IoNTs) a reality. Neural communication has been proposed as a promising solution, utilizing action potentials (APs) as the fundamental units for information transmission. Most of the existing research simplifies the membrane potential into two states: an excited state that generates an AP upon stimulation and a resting state absent of stimulation. This assumption neglects both the intrinsic oscillation of membrane potential and the uncertainty it introduces into information transmission. Therefore, neural communication requires further investigation into the biological similarity of channel models and the reliability of transmission. In this paper, we employ the Izhikevich model as the channel model to characterize membrane potential oscillations. Additionally, we devise an enhanced code division multiplexing (CDM) scheme based on this model, enabling multiple signals to share a single neuron channel. In contrast to previous CDM methods, this scheme employs further encoding of superimposed signals. The performance is evaluated in terms of bit error rate (BER), and the results indicate a significant improvement in interference resistance. This research improves the communication efficiency of engineered neural systems and achieves more precise and reliable communication.