Adaptive deep brain stimulation (aDBS) has enabled machine learning models to track motor states from neural signals with improved accuracy, aiming to provide electrical stimulation accordingly. Such data-driven techniques necessitate extensive user-specific, synchronized kinematic and neural data collection involving repetitive tasks and additional sensors to quantify continuous movements, due to variations in neural signals among individuals. In this study, we introduce the Neural-to-Gait Calibrator Framework (N2GCF), a cross-subject deep learning framework that leverages collective neural data to track gait performance of users with Parkinson's disease. Our framework utilizes domain adversarial learning to map a target user's unlabeled neural features with labeled data from other individuals, removing the need for synchronous gait recording systems thereby allowing personalized model calibration outside equipped clinical settings. The framework's ability was demonstrated through a significant reduction in error rates compared to models trained with data from other individuals without calibration, with its architecture achieving performance comparable to that of models trained directly with labeled target data. Our N2GCF provides an effective biomarker reflecting gait performance for aDBS systems despite being trained with only unlabeled neural signals for the target, enabling the model to be established in any location where stepping in place can be performed.