Channel Frequency Response Estimation Using Machine Learning

The patent applies preliminary action by pre-training a machine learning model using comprehensive channel response data across multiple frequency spectrums. This pre-training phase performs the resource-intensive measurements and data collection in advance, creating a trained model that can later rapidly estimate channel responses without requiring repeated full-scale measurements. The model learns from historical channel characteristics and can predict performance for new technologies quickly.

The patent uses copying by creating a virtual representation of the channel through machine learning model predictions rather than performing physical measurements. The trained model generates estimated channel frequency responses that replicate the characteristics of actual measurements, allowing multiple scenarios to be evaluated by copying and applying the same model to different frequency spectrums and technology configurations without repeating the original measurement process.

The patent applies universality by creating a single machine learning model that can predict channel performance across multiple frequency spectrums and for different communication technologies. The model is trained on diverse channel response data and can universally apply learned patterns to estimate performance for various scenarios, eliminating the need to perform separate comprehensive tests for each technology or frequency spectrum.

The patent uses parameter changes by transforming the approach from direct physical measurement to machine learning-based estimation. The system changes the fundamental parameter of how channel response is obtained - instead of measuring actual physical channels for each scenario, it uses a trained model that processes input parameters (frequency spectrum, technology type) to generate predicted channel responses, significantly reducing resource requirements.

The patent uses an intermediary approach by introducing a machine learning model as a mediator between narrowband measurements and wideband predictions. The model acts as an intermediary that takes simple narrowband measurement data as input and transforms it into comprehensive wideband channel response estimates, bridging the gap between simplified measurements and the need for broad frequency spectrum coverage.

The patent applies dimensionality change by expanding the frequency spectrum dimension through machine learning. Instead of directly measuring across the entire wideband spectrum, the system performs measurements in the narrowband dimension, then uses the trained model to extrapolate and generate channel response characteristics across the wider frequency spectrum dimension, effectively adding spectral coverage through computational methods.