A Novel Combination of Neural Networks and FFT for Frequency Estimation of SAW Resonators’ Responses

The surface acoustic wave (SAW) resonator sensor transmits its damped resonant responses to the reader (interrogation unit). The interrogation unit has the responsibility of detecting this resonant frequency in a short time in order to find the amount of the measurand. The more accurately the frequency is estimated, the higher resolution the measurement can possess. The Discrete Fourier Transform (DFT) can demonstrate the resonant frequency where power of the response signal is the maximum. By employing approximations such as zero padding, parabolic approximation, and rectangular windowing, the accuracy of the DFT can increase well. However, the DFT and its approximations can be affected by noise of the channel and the high frequency blocks of the reader in a sensitive manner. This paper presents a noise-resistant frequency estimation method based on neural networks in order to achieve a high accuracy in harsh environments or for the long channels. Measurement results of an experimental setup are presented and frequency detection methods are implemented on a microcontroller to validate its efficiency and reliability.