SignalBuilder

The simulation of temporal signals is crucial in fields such as deep learning and signal processing, where access to real data is often limited due to ethical, economic, or privacy concerns. This study presents a method for generating synthetic data by creating frequency reference points within controlled intervals, enabling the modeling of real phenomena with high precision. The generated signals are complex and variable, capturing both smooth transitions and abrupt changes, making them valuable tools for evaluating the effectiveness of advanced reconstruction models, such as Generative Adversarial Networks (GANs). The approach allows for the creation of signals at different resolutions through oversampling and undersampling, tailoring the level of detail to the needs of the analysis. Additionally, parameters such as amplitude, phase, frequency, and noise levels can be adjusted to simulate realistic conditions. These synthetic data provide a controlled and reproducible environment for algorithm evaluation, with applications in fields like biomedicine, telecommunications, and clinical data analysis.