Deep Convolutional Neural Network Architectures for Tonal Frequency Identification in a Lofargram
- PDF / 1,157,742 Bytes
- 10 Pages / 594.77 x 793.026 pts Page_size
- 38 Downloads / 257 Views
ISSN:1598-6446 eISSN:2005-4092 http://www.springer.com/12555
Deep Convolutional Neural Network Architectures for Tonal Frequency Identification in a Lofargram Jihun Park* and Dae-Jin Jung Abstract: Advances in convolutional neural networks (CNNs) have driven the development of computer vision. Recent CNN architectures, such as those with skip residual connections (ResNets) or densely connected architectures (DenseNets), have facilitated backpropagation and improved the performance of feature extraction and classification. Detecting objects in underwater environments by analyzing sound navigation and ranging (sonar) signals is considered an important process that should be automated. Several previous approaches have addressed this challenge; however, there has been no in-depth study of CNN architectures that effectively analyze sonar grams. In this paper, we have presented the identification of tonal frequencies in lofargrams using recent CNN architectures. Our study includes 175 CNN models that are derived from five different CNN architectures and 35 different input patch sizes. The study results showed that the accuracy of the best model was as high as 96.2% for precision and 99.5% for recall, with an inference time of 0.184 s. Keywords: Convolutional neural networks, lofar analysis, sonar analysis, underwater recognition.
1.
INTRODUCTION
In an underwater environment, it is difficult to visually identify objects because the visual range is much lower than in air. Sound, which travels four times faster in water than in air, has been widely used to detect objects underwater. Sound navigation and ranging (sonar) is a technique that uses sound to navigate or detect objects underwater. Sonar can be classified into active sonar and passive sonar. Active sonar generates specific pulses of sounds, and then detects objects by using the reflections of the emitted sound. Passive sonar only listens to sound from other objects, especially unknown ships or submarines. In the military domain, passive sonar is more useful than active sonar because sound transmission can reveal the location of a ship. When a ship or a submarine is maneuvered, the rotation of shafts and blades make a specific sound pulse. Even when they remain in place, auxiliary devices such as air conditioning systems or power systems, can emit sound. A passive sonar signal detection system analyzes these sounds. An array of hydrophones is used to reduce noise and to increase the level of the actual sound. The sound obtained by hydrophones is processed using signal processing methods to reduce background noise and extract the signal features of a specific ship or submarine. Low frequency analysis and recording (LOFAR) [1] and
detection envelope modulation on noise (DEMON) [2] are frequently used signal processing methods in passive sonar signal detection [3]. LOFAR and DEMON transform sound into images, which can help sonar technicians visually recognize signal features. Sonar technicians listen to hydrophones and analyze the signals to recognize unknown shi
Data Loading...
