A waveform-based convolutional neural net for source range and ocean environment classification

Abstract

Neural networks learn features that are useful for classification directly from a source, such as a recorded signal, which removes the need for feature extraction or domain transformations necessary in other machine learning algorithms. To take advantage of these benefits and have a finer temporal resolution than a spectrogram, we built a one-dimensional convolutional neural network to classify source range and ocean environment from a received signal. The neural network was trained on simulated signals generated in different environments (sandy, muddy, or mixed-layer sediment layers) for several range classes. We found significant potential in a neural network of this type, given a large amount of varied training samples for the network to learn important features to make range and environment predictions. This type of network provides an alternative for frequency-domain learning and is potentially useful for impulsive sources. Benefits of using a time-domain envelope are also explored. Success in the time domain also reduces the computational requirements of conversion to frequency domain and increases the temporal resolution, which might be beneficial for real-time applications. [Work supported by the Office of Naval Research.] Neural networks learn features that are useful for classification directly from a source, such as a recorded signal, which removes the need for feature extraction or domain transformations necessary in other machine learning algorithms. To take advantage of these benefits and have a finer temporal resolution than a spectrogram, we built a one-dimensional convolutional neural network to classify source range and ocean environment from a received signal. The neural network was trained on simulated signals generated in different environments (sandy, muddy, or mixed-layer sediment layers) for several range classes. We found significant potential in a neural network of this type, given a large amount of varied training samples for the network to learn important features to make range and environment predictions. This type of network provides an alternative for frequency-domain learning and is potentially useful for impulsive sources. Benefits of using a time-domain envelope are also explored. Success in the ti...