Indoor Positioning Based on Frequency Fading Characteristics of Wideband Signals

As location-based services in indoor environment have burgeoned, there has been a corresponding surge in research on indoor positioning technologies. Current methods, like Channel State Information (CSI) fingerprinting, impose rigorous requirements on receiving devices, mandating specialized CSI acquisition hardware and intricate data collection software. This paper introduces a novel indoor positioning algorithm leveraging the frequency fading characteristics (FFC) of wideband signals, implementable using a single-antenna apparatus for direct data capture. The initial step involves processing raw data through the Fast Fourier Transform (FFT) to obtain the FFC within the frequency domain. Subsequently, the frequency spectrum of the signals of interest was extracted. A convolutional auto-encoder (CAE) network is then harnessed to distill features from the spectrogram, primarily aiming to curtail data size, mitigate computational demands, and economize memory consumption. In the concluding step, a Residual Network (ResNet) model is adopted to categorize these features, establishing a correlation between the features and reference points. Experimental validation in a conference room setup yielded a positioning precision of 0.67m. The accuracy and robustness of the proposed algorithm outshine conventional indoor positioning approaches, including Received Signal Strength (RSS) and Angle of Arrival (AOA) positioning.