The fast expansion of Internet of Things (IoT) networks operating in healthcare settings and low-power environments requires robust communication systems which maintaintheir performance in sparse and noisy communication channels. The performance of OTFS modulation becomes worse when it operates under conditions of high sparsity and elevated noise levels. The current fixed-transform methods, which use wavelet, DCT, and spline techniques, do not adapt to non-stationary signal patterns, which limits their ability to function well in dynamic IoT environments. The proposed HATS framework (Hybrid Adaptive Transform Selection) selects the most suitable transform domain dynamically to boost OTFS transmission by adapting to signal sparsity and noise characteristics. The evaluation of HATS on the MIT-BIH Arrhythmia Database for ECG signal recovery and QRS detection resulted in a minimum MSE of 0.000268, 41.57 dB PSNR, and R-peak correlation, which surpassed fixed-transform baselines. The proposed improvements become vital for biomedical IoT applications because they maintain diagnostic feature preservation of QRS complexes and reduce energy consumption. The HATS framework shows effective capabilities to boost reliability in IoT systems which handle restricted data while dealing with communication interference.