Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Frequency Tracking Features for Data-Efficient Deep Siren Identification

Stefano Damiano, Thomas Dietzen, Toon van Waterschoot

TL;DR

This work proposes a low-complexity feature extraction method based on frequency tracking using a single-parameter adaptive notch filter that consistently outperforms the traditional spectrogram-based model when limited training data is available, achieves better cross-domain generalization and has a smaller size.

Abstract

The identification of siren sounds in urban soundscapes is a crucial safety aspect for smart vehicles and has been widely addressed by means of neural networks that ensure robustness to both the diversity of siren signals and the strong and unstructured background noise characterizing traffic. Convolutional neural networks analyzing spectrogram features of incoming signals achieve state-of-the-art performance when enough training data capturing the diversity of the target acoustic scenes is available. In practice, data is usually limited and algorithms should be robust to adapt to unseen acoustic conditions without requiring extensive datasets for re-training. In this work, given the harmonic nature of siren signals, characterized by a periodically evolving fundamental frequency, we propose a low-complexity feature extraction method based on frequency tracking using a single-parameter adaptive notch filter. The features are then used to design a small-scale convolutional network suitable for training with limited data. The evaluation results indicate that the proposed model consistently outperforms the traditional spectrogram-based model when limited training data is available, achieves better cross-domain generalization and has a smaller size.

Frequency Tracking Features for Data-Efficient Deep Siren Identification

TL;DR

This work proposes a low-complexity feature extraction method based on frequency tracking using a single-parameter adaptive notch filter that consistently outperforms the traditional spectrogram-based model when limited training data is available, achieves better cross-domain generalization and has a smaller size.

Abstract

The identification of siren sounds in urban soundscapes is a crucial safety aspect for smart vehicles and has been widely addressed by means of neural networks that ensure robustness to both the diversity of siren signals and the strong and unstructured background noise characterizing traffic. Convolutional neural networks analyzing spectrogram features of incoming signals achieve state-of-the-art performance when enough training data capturing the diversity of the target acoustic scenes is available. In practice, data is usually limited and algorithms should be robust to adapt to unseen acoustic conditions without requiring extensive datasets for re-training. In this work, given the harmonic nature of siren signals, characterized by a periodically evolving fundamental frequency, we propose a low-complexity feature extraction method based on frequency tracking using a single-parameter adaptive notch filter. The features are then used to design a small-scale convolutional network suitable for training with limited data. The evaluation results indicate that the proposed model consistently outperforms the traditional spectrogram-based model when limited training data is available, achieves better cross-domain generalization and has a smaller size.
Paper Structure (5 sections, 13 equations, 2 figures, 3 tables, 1 algorithm)

This paper contains 5 sections, 13 equations, 2 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem Statement and Baseline
  3. Proposed Method
  4. Evaluation
  5. Conclusions

Figures (2)

  • Figure 1: Proposed features for three audio samples: frequency tracked by the ANF algorithm (above, highlighted in white and overlaid to the full spectrogram) and power ratio (below).
  • Figure 2: Comparison of the average (solid line) and standard deviation (shaded area) of the F1-score for the baseline VGGSiren and the proposed ANFNet, trained with an increasing amount of data: in-domain evaluation (above) and cross-dataset evaluation (below).