Spiking Neural Networks (SNNs) offer energy-efficient computation, but conventional training methods focus on optimizing synaptic weights using homogeneous neurons, neglecting neuronal diversity. This limitation restricts the learning capacity of SNNs and leads to inefficiencies, such as excessive firing rates from high input voltages or poor information representation from low voltages. To address these issues, we propose a novel spiking neuron based on Leaky-Integrate-Fire (LIF) model, which is equipped with Learnable firing Threshold and input Resistance, termed LTR-LIF, enabling optimization of dynamic parameters in directly trained SNNs. Through simulations, we demonstrate that adjusting these parameters improves neuronal adaptability and network performance. Additionally, we introduce the Adaptive Sparse Loss (ASL) function to regulate firing rates, promoting sparsity and energy efficiency. Evaluated on static and neuromorphic datasets, our approach achieves state-of-the-art accuracy—96.25% on CIFAR-10, 81.03% on CIFAR-100, and 60.33% on Tiny-ImageNet.