Automatic analysis of audiostreams in the concept drift environment

CAAI Trans on Intel Tech

Potamitis

2021

Self Cite

Computational bioacoustics is a relatively young research area, yet it has increasingly received attention over the last decade because it can be used in a wide range of applications in a cost-effective manner. This work focuses on the problem of detecting the novel bird calls and songs associated with various species and individual birds. To this end, variational autoencoders, consisting of deep encoding-decoding networks, are employed. The encoder encompasses a series of convolutional layers leading to a smooth high-level abstraction of log-Mel spectrograms that characterise bird vocalisations. The decoder operates on this latent representation to generate each respective original observation. Novel species/individual detection is carried out by monitoring and thresholding the expected reconstruction probability. We thoroughly evaluate the proposed method on two different data sets, including the vocalisations of 11 North American bird species and 16 Athene noctua individuals. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

Section: Figures Of Merit and Parameterisationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Acoustic detection of unknown bird species and individuals

CAAI Trans on Intel Tech

Potamitis

2021

Self Cite

“…The proposed algorithm belongs to passive concept drift detection class [22], [23] since it doesn't actively seek changes in the environment, unlike for example approaches such as [24]. It relies on user's input indicating the appearance of a new class via the mobile application.…”

Section: Algorithm To Update Class Dictionary Based On User's Inputmentioning

confidence: 99%

Acoustic classification of individual cat vocalizations in evolving environments

2021 44th International Conference on Telecommunications and Signal Processing (TSP)

Sánchez²

2021

Self Cite

This paper is focused on the classification of vocalizations characterizing intents of individual cats. Cats vocalize in order to convey different emotions and/or intents and although their repertoire/vocabulary may not be universal, it exhibits consistent characteristics on an individual basis. In this work, we present a complete pipeline for processing streams of audio, with the twofold goal being both detection as well as interpretation of cat vocalizations. The proposed system is based on YAMNet pre-trained deep network where we apply meaningful modifications addressing the requirements of the task-athand. Interestingly, the overall system is able to run in real-time on modern smartphones using the developed application. At the same time, we address the non-stationarity problem meaning that class dictionary is updated on-the-fly following user recommendations. After extensive experiments, we show that the proposed pipeline achieves quite satisfactory detection and recognition accuracy. To the best of our knowledge, this is the first attempt in the related literature to address continuous detection and interpretation of cat vocalizations in a real-time fashion.

“…MFCCs comprise a short-term power spectrum signal representation, where the frequency bands are distributed according to a Mel-scale instead of the linearly-spaced approach. This type of feature is widely used in the field of audio analysis due to its discriminating power (Ntalampiras, 2016). The Melscale filter bank maps the powers of the spectrum using triangular overlapping windows.…”

Section: Audio Featuresmentioning

confidence: 99%

Automatic Acoustic Diagnosis of Heartbeats

Mastrangelo

Proceedings of the 18th International Conference on Signal Processing and Multimedia Applications

2021

Self Cite

Automatic identification of heart irregularities based on the respective acoustic emissions is a relevant research field which receives ever-increasing attention over the last years. Devices such as digital stethoscope and smartphones can record the heartbeat sounds and are easily accessible, making this method more appealing. This paper presents different automatic procedures to classify heartbeat sounds coming from such devices into five different labels: normal, murmur, extra heart sound, extrasystole and artifact so that even people without medical knowledge can detect heart irregularities. The data used in this paper come from two different datasets. The first dataset is collected through an iPhone application whereas the second one is collected from a digital stethoscope. To be able to classify heartbeat sounds, time and frequency domain features are extracted and modeled by different machine learning algorithms, i.e. k-NN, random forest, SVM and ANNs. We report the achieved performances and a thorough comparison.