A Novel Underwater Location Beacon Signal Detection Method Based on Mixing and Normalizing Stochastic Resonance

Guolong Liang, Guangming Wan, Jinjin Wang, Xue WangSensors 2020 · DOI

TL;DR

This paper introduces a novel method called Mixing and Normalizing Stochastic Resonance (MNSR) to detect weak underwater locator beacon (ULB) signals, which are notoriously difficult to pick up due to their high frequency and short pulse duration. MNSR overcomes the limitations of traditional stochastic resonance by effectively down-mixing the signal and normalizing its parameters, enabling detection even with lower sampling rates and shorter signal durations.

What Problem Does This Paper Address?

When an aircraft crashes into water, the underwater locator beacon (ULB) on the flight data recorder (FDR) is crucial for locating the wreckage. These ULBs emit a 37.5 kHz ultrasonic pulse, but due to their small size, limited battery, and the significant signal attenuation in water, these signals become extremely weak and difficult to detect beyond a few kilometers. Conventional detection methods struggle with the low signal-to-noise ratio (SNR). Stochastic Resonance (SR) offers a promising avenue for weak signal detection by leveraging noise to amplify the signal. However, classical SR has strict limitations: it requires the input signal to be "small-parameter" (low amplitude and frequency) and a high sampling frequency (at least 50 times the signal frequency). These conditions are not met by ULB signals, necessitating a new approach.

Methods

The paper proposes a novel method called Mixing and Normalizing Stochastic Resonance (MNSR). It's a three-stage process:

  1. Mixing: The weak ULB signal (with frequency $f$) is mixed with a carrier signal (frequency $f_c$). This process generates a new signal containing a difference frequency component ($|\text{f}_c - f|$) and a sum frequency component. Crucially, the difference frequency can be made much lower than the original ULB signal frequency. This down-mixing is key to overcoming the high-frequency limitation of classical SR.
  2. Low-Pass Filtering: A low-pass filter is applied to remove the sum frequency component, leaving only the down-mixed difference frequency signal and modulated noise.
  3. Normalizing Stochastic Resonance (PNSR): The filtered signal is then processed using a form of SR (specifically, PNSR is mentioned, which normalizes system parameters). This stage amplifies the weak difference frequency signal.

A significant contribution is the parameter adjustment method for MNSR. This method guides how to select the carrier frequency ($f_c$), cutoff frequency ($f_{cut}$), and barrier parameters ($a, b$) to effectively implement MNSR, especially when dealing with large-parameter signals and lower sampling frequencies.

Datasets

The paper does not explicitly mention or introduce new datasets. Instead, it relies on:

  • Simulated ULB signals: Generated based on the known characteristics of ULB signals (frequency, pulse width, source level) and underwater propagation models.
  • Experimental data: Collected in an anechoic tank using a 37.5 kHz acoustic source to simulate a ULB signal. The data acquisition system used a sampling frequency of 150 kHz.

Key Insights and Ideas

  • Down-mixing is the key: By mixing the high-frequency ULB signal with a carrier, the problem is transformed into detecting a lower-frequency difference signal. This bypasses the direct high-frequency limitation of classical SR.
  • Relaxing the sampling rate constraint: MNSR significantly reduces the requirement for extremely high sampling frequencies (50x signal frequency). The paper shows it can work effectively even when the sampling frequency is much lower than this threshold, making it more practical.
  • Handling large-parameter signals: Unlike classical SR, MNSR is designed to handle signals with large amplitudes and frequencies, which are typical for ULB signals.
  • Practical parameter tuning: The proposed parameter adjustment method is crucial. It provides a systematic way to configure MNSR for specific ULB signal characteristics and noise conditions, making the method applicable in real-world scenarios.
  • Experimental validation: The successful demonstration of MNSR in a controlled tank experiment validates its practical utility beyond numerical simulations.

Key Results

  • Numerical Simulations:
    • MNSR successfully detected a simulated ULB signal (37.5 kHz, low SNR) where classical SR failed due to large-parameter limitations.
  • The method showed effectiveness even with a sampling frequency ratio ($k = f_s/f$) as low as 4, far below the classical SR requirement of $k \ge 50$.
  • Compared to RFSR (Rescaling Frequency Stochastic Resonance), MNSR achieved detection with a significantly lower sampling frequency and shorter signal duration.
  • Compared to MSR (Modulated Stochastic Resonance), MNSR was effective for shorter pulse durations, which is critical for ULB signals.
  • Experimental Results:
    • In a tank experiment simulating a ULB signal, MNSR successfully identified and enhanced the signal, which was barely discernible in the raw input. A distinct spectral peak at the down-mixed frequency was observed in the MNSR output.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that classical SR has limitations regarding small parameters and high sampling frequencies. MNSR is presented as a solution to these specific limitations.
  • The effectiveness of MNSR is demonstrated for a specific ULB signal frequency (37.5 kHz) and pulse characteristics. Its performance across a wider range of ULB signal parameters and diverse underwater noise environments might require further investigation.
  • While the parameter adjustment method is provided, fine-tuning these parameters for optimal performance in highly variable real-world ocean conditions could be challenging.
  • Future Work:
    • The authors suggest that MNSR can be an "active approach in practical weak signal detection scenes," implying its potential for deployment in actual search and rescue operations.
  • Further research could focus on optimizing the parameter adjustment method for real-time adaptation to changing environmental conditions and noise levels.
  • Exploring the application of MNSR to other types of weak underwater acoustic signals beyond ULBs could be a valuable extension.

Summary updated

Abstract

A flight data recorder (FDR) is an electronic recording device placed in an aircraft for the purpose of facilitating the investigation of aviation accidents. If an aircraft crashes into water, an underwater locator beacon (ULB), which is installed on the FDR, is triggered by water immersion, and emits an ultrasonic 10 ms pulse signal once per second at 37.5 kHz. This pulse signal can be detected by sonar equipment. However, the ULB signal only can be detectable 1-2 kilometers from the surface in normal conditions. Stochastic resonance (SR) is a rising theory in the field of weak signal detection. The classical stochastic resonance limits state that the input must be small-parameter and the sampling frequency must be 50 times higher than the signal frequency. It cannot be applied to the ULB signal detection. To resolve this problem, this paper presents a novel approach named mixing and normalizing stochastic resonance (MNSR). By mixing the ULB signal and normalizing SR system parameters, MNSR provides a new way to detect weak ULB signal. Meanwhile, we propose the parameters adjustment method of MNSR. We prove the effectiveness through numerical simulation. An experiment in a tank is employed to verify the practicability of this method.
BibTeX
@inproceedings{liang2020novel,
  author = {Guolong Liang and Guangming Wan and Jinjin Wang and Xue Wang},
  title = {{A Novel Underwater Location Beacon Signal Detection Method Based on Mixing and Normalizing Stochastic Resonance}},
  booktitle = {Sensors},
  year = {2020},
  doi = {10.3390/s20051292},
}
Figure