Machine learning algorithms have shown their abilities to tackle difficult recognition problems, but they are still rife with challenges. Among these challenges is how to deal with problems where new categories constantly occur, and the datasets can dynamically grow. Most contemporary learning algorithms developed to this point are governed by the assumptions that all testing data classes must be the same as training data classes, often with equal distribution. Under these assumptions, machine-learning algorithms can perform very well, using their ability to handle large feature spaces and classify outliers. The systems under these assumptions are called Closed Set Recognition systems (CSR). However, these assumptions cannot reflect practical applications in which out-of-set data may be encountered. This adversely affects the recognition prediction performances. When samples from a new class occur, they will be classified as one of the known classes. Even if this sample is far from any of the training samples, the algorithm may classify it with a high probability, that is, the algorithm will not only be wrong, but it may also be very confident in its results. A more practical problem is Open Set Recognition (OSR), where samples of classes not seen during training may show up at testing time. Inherently, there is a problem how the system can identify the novel sound classes and how the system can update its models with new classes. This thesis highlights the problems of multi-class recognition for OSR of sounds as well as incremental model adaptation and proposes solutions towards addressing these problems. The proposed solutions are validated through extensive experiments and are shown to provide improved performance over a wide range of openness values for sound classification scenarios.

When a boat moves in water, it creates a sound with unique features which makes it possible to identify different boat types or even a specific boat. The ability to identify boats is important in the military sector for surveillance purposes.This thesis describes how different audio processing methods and machine learning approaches are implemented, tested and evaluated in order to create a prototype that identifies boats. A total of 87 boat sounds were used and processed in seven different ways. The machine learning approaches Dense Neural Network, Convolutional Neural Network and Recurrent Neural Network were implemented and trained with the processed audio files in order to identify different boat types. Different combinations of audio processing methods and machine learning approaches ability to classify different boat types, were tested with a stratified Kfold test.The result is a prototype with an audio processing method that divides an audio file to equally large segments. Each segment is converted to a logarithmic mel-scaled spectrogram and a delta feature is calculated and added as an extra dimension for each segment. A Convolutional Neural Network is trained with processed audio files and manages to distinguish different boat types with an accuracy of 75%. / En båt kan identifieras genom att analysera ljudet den skapar när den rör sig i vatten. Förmågan att identifiera båtar är viktig ur övervakningssynpunkt i den militära sektorn. Den här rapporten beskriver hur olika metoder inom ljudanalys och maskininlärning har implementerats, testats och utvärderats för att skapa en prototyp som kan identifiera olika båtar. Totalt 87 olika båtljud användes och behandlades på sju olika sätt.Inom området maskininlärning användes teknikerna ”Dense Neural Network”, ”Convolutional Neural Network” och ”Recurrent Neural Network” som tränades för att identifiera olika båttyper. Olika kombinationer av metoder inom ljudbehandling och maskininlärning testades med ett ”stratified Kfold” test för att utvärdera förmågan att klassificera olika båttyper.Resultatet blev en prototyp med en ljudbehandlingsmetod som delar upp en ljudfil i segment av samma storlek. Varje segment konverteras till ett ”logaritmiskt mel-scaled spectrogram” och en extra dimension med ett deltavärde adderas. Ett ” Convolutional Neural Network” tränas med de behandlade ljudfilerna och lyckas urskilja olika båtklasser med 75% sannolikhet.