Wavelets in Recognition of Bird Sounds

Abstract

This paper presents a novel method to recognize inharmonic and transient bird sounds efficiently. The recognition algorithm consists of feature extraction using wavelet decomposition and recognition using either supervised or unsupervised classifier. The proposed method was tested on sounds of eight bird species of which five species have inharmonic sounds and three reference species have harmonic sounds. Inharmonic sounds are not well matched to the conventional spectral analysis methods, because the spectral domain does not include any visible trajectories that computer can track and identify. Thus, the wavelet analysis was selected due to its ability to preserve both frequency and temporal information, and its ability to analyze signals which contain discontinuities and sharp spikes. The shift invariant feature vectors calculated from the wavelet coefficients were used as inputs of two neural networks: the unsupervised self-organizing map (SOM) and the supervised multilayer perceptron (MLP). The results were encouraging: the SOM network recognized 78% and the MLP network 96% of the test sounds correctly.

Reconocimiento automatizado de señales bioacústicas: Una revisión de métodos y aplicaciones

“…Algunos trabajos en bioacústica resaltan sus propiedades en la representación de señales no estacionarias, con energía dispersa en un amplio rango de frecuencias, discontinuas y con picos de energía [10], [48], [49].…”

A Syllable-Level Probabilistic Framework for Bird Species Identification

“…These feature vectors then act as inputs to a classification algorithm for identifying the particular bird species. Different feature representations and machine learning methods have been applied for bird species identification in the literature [1], [3]- [6]. In [4], the authors used dynamic time warping (DTW) to compare the input spectrograms with a predefined set of templates.…”

“…In [5], the authors used neural networks and multivariate statistical techniques in conjunction with a set of temporal and spectral features. In [6], the authors used wavelet coefficients along with self organizing map (SOM) and multilayer perceptron (MLP). In [1], the authors compared three different feature representations (sinusoidal model, Mel-cepstrum model, descriptive parameters) by evaluating their performance with different classification algorithms based on DTW, Gaussian mixture model (GMM), Hidden Markov model (HMM).…”

Acoustic classification of Australian frogs for ecosystem survey

“…WPD is a powerful tool for the analysis of non-stationary signals, which includes multiple bases and different basis [Selin et al, 2007]. With WPD, an original acoustic signal can be split into two frequency bands such as lower and higher frequency band.…”

“…Then, both lower and higher frequency bands can be further continuously decomposed into two sub-bands, which produce a complete wavelet packet tree [Farooq and Datta, 2001]. Due to its ability for analysing a nonstationary signal, WPD has been used to analyse acoustic signals [Ren et al, 2008, Selin et al, 2007. Here, WPD is used to obtain features for frog call classification.…”