Bird sound spectrogram decomposition through Non-Negative Matrix Factorization for the acoustic classification of bird species

Abstract

Feature extraction for Acoustic Bird Species Classification (ABSC) tasks has traditionally been based on parametric representations that were specifically developed for speech signals, such as Mel Frequency Cepstral Coefficients (MFCC). However, the discrimination capabilities of these features for ABSC could be enhanced by accounting for the vocal production mechanisms of birds, and, in particular, the spectro-temporal structure of bird sounds. In this paper, a new front-end for ABSC is proposed that incorporates this specific information through the non-negative decomposition of bird sound spectrograms. It consists of the following two different stages: short-time feature extraction and temporal feature integration. In the first stage, which aims at providing a better spectral representation of bird sounds on a frame-by-frame basis, two methods are evaluated. In the first method, cepstral-like features (NMF_CC) are extracted by using a filter bank that is automatically learned by means of the application of Non-Negative Matrix Factorization (NMF) on bird audio spectrograms. In the second method, the features are directly derived from the activation coefficients of the spectrogram decomposition as performed through NMF (H_CC). The second stage summarizes the most relevant information contained in the short-time features by computing several statistical measures over long segments. The experiments show that the use of NMF_CC and H_CC in conjunction with temporal integration significantly improves the performance of a Support Vector Machine (SVM)-based ABSC system with respect to conventional MFCC.

Bioacoustic classification of avian calls from raw sound waveforms with an open-source deep learning architecture

“…Processing bioacoustic data has typically relied on the initial extraction of features from the audio signal, which are often processed further into handcrafted features prior to their use in machine learning. Researchers in the field are aware of the limitations of this approach, as the most widely used handcrafted features are designed for human speech tasks and focus on spectral characteristics of the sound that may differ from the intended bioacoustic tasks 18 , 19 . However, traditional deep learning techniques applied directly to the raw waveform have remained largely out of reach, due to limitations primarily associated with insufficient volumes of labelled data.…”

“…Labelled sounds in a training dataset allow the algorithm to learn and make predictions for new sounds 17 . Researchers have used a variety of machine learning algorithms such as Hidden Markov Models, Gaussian Mixture Models and Support Vector Machines 18 . More recently the focus has shifted toward the use of deep learning methods such as Convolutional Neural Networks (CNN) 7 .…”

“…These may be further processed into handcrafted features, such as Mel frequency cepstral coefficients, that transform the linear frequency according to a perceptual scale (Mel scale) reflecting the non-linear human perception of sound 20 . The popular use and relative success of the Mel scale in bioacoustics is being progressively challenged by researchers, as the use of the acoustic spectrum by other animals is likely to differ from that derived from human perception 18 , 19 .…”

“…human speech processing) and their performance may vary between tasks 1 . Extracted feature selection is critical for performance and may involve trial and error, some level of automation 18 , or may just follow previous successes possibly impacting performance when applied to new tasks or datasets. Training directly from the raw waveform bypasses this step, allowing the algorithm to select automatically the elements of the sound best suited for the task that could otherwise be lost at the feature selection or extraction step 23 .…”

Multi-Level Local Feature Coding Fusion for Music Genre Recognition

“…The non-negative matrix factorization (NMF) [31], the non-negative tensor factorization (NTF) [32], and the sparse coding [33] are instances of unsupervised learning methods for MGR. In addition to the sparse coding, there are many feature coding methods used in image classification, including the hard coding [34], the soft coding [35], the low-rank sparse coding [36] [37], the vector of locally aggregated descriptor (VLAD) coding [18], and the Fisher vector (FV) coding [38].…”