Automatic large-scale classification of bird sounds is strongly improved by unsupervised feature learning

By Dan Stowell, Mark D. Plumbley. PeerJ, 2014

295 citations recorded by the source corpus.

Automatic species classification of birds from their sound is a computational tool of increasing importance in ecology, conservation monitoring and vocal communication studies. To make classification useful in practice, it is crucial to improve its accuracy while ensuring that it can run at big data scales. Many approaches use acoustic measures based on spectrogram-type data, such as the Mel-frequency cepstral coefficient (MFCC) features which represent a manually-designed summary of spectral information. However, recent work in machine learning has demonstrated that features learnt automatically from data can often outperform manually-designed feature transforms. Feature learning can be performed at large scale and “unsupervised”, meaning it requires no manual data labelling, yet it can improve performance on “supervised” tasks such as classification. In this work we introduce a technique for feature learning from large volumes of bird sound recordings, inspired by techniques that have proven useful in other domains. We experimentally compare twelve different feature representations derived from the Mel spectrum (of which six use this technique), using four large and diverse databases of bird vocalisations, classified using a random forest classifier. We demonstrate that in our classification tasks, MFCCs can often lead to worse performance than the raw Mel spectral data from which they are derived. Conversely, we demonstrate that unsupervised feature learning provides a substantial boost over MFCCs and Mel spectra without adding computational complexity after the model has been trained. The boost is particularly notable for single-label classification tasks at large scale. The spectro-temporal activations learned through our procedure resemble spectro-temporal receptive fields calculated from avian primary auditory forebrain. However, for one of our datasets, which contains substantial audio data but few annotations, increased performance is not discernible. We study the interaction between dataset characteristics and choice of feature representation through further empirical analysis.

Subfields: Developmental Biology, Ecology, Signal Processing

View Paper: https://www.semanticscholar.org/paper/62b8168eb6952e23998fc31ca4ce240973de1ce3

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Developmental Biology

In Ecology

In Signal Processing

Citations in the corpus

Finding, visualizing, and quantifying latent structure across diverse animal vocal repertoires (2020, methodology): “Mel frequency scaling was used as it has previously proven useful in extracting features of animal vocalizations [111], although in some cases linear frequency scaling can perform better in bioacoustics [86].”
Machine learning in acoustics: Theory and applications. (2019, background): “Stowell and Plumbley 208 provide an example of using a spherical K-means learner to construct features from Mel-ﬁltered spectra.”
Detection and Classification of Acoustic Scenes and Events: Outcome of the DCASE 2016 Challenge (2018, methodology): “current studies, methods investigated include GMMs [30]–[32], SVMs combined with multiple instance learning [33], unsupervised feature learning [34], [35] and CNNs [36].”
A Closer Look at Weak Label Learning for Audio Events (2018, background): “Human-computer interaction [18], smart devices including voice-based assistants 5, smart homes and health monitoring systems [19][20][21], [22], wildlife monitoring [23], all require automated audio content”
End-to-End Environmental Sound Classification using a 1D Convolutional Neural Network (2019, background): “Furthermore, 1D representations are noisier compared to 2D representations (Stowell & Plumbley, 2014).”
Emerging opportunities and challenges for passive acoustics in ecological assessment and monitoring (2018, methodology): “…test for new methods, and the involvement of the machine learning and computer vision communities in PAM is driving analytical advances that benefit ecologists (Goeau, Glotin, Vellinga, Planque, & Joly, 2016; Marinexplore, 2013; Stowell & Plumbley, 2014; Stowell, Wood, Stylianou, & Glotin, 2016).”
Automatic acoustic detection of birds through deep learning: The first Bird Audio Detection challenge (2018, methodology): “Python code for each of the baseline classifiers has previously been published (Stowell & Plumbley, 2014a; Stowell et al., 2015).”
Data-efficient weakly supervised learning for low-resource audio event detection using deep learning (2018, background): “Methods with fewer limitations were introduced in more recent studies [21, 5, 4, 23, 34].”