Time-frequency segmentation of bird song in noisy acoustic environments

Abstract

Recent work in machine learning considers the problem of identifying bird species from an audio recording. Most methods require segmentation to isolate each syllable of bird call in input audio. Energy-based time-domain segmentation has been successfully applied to low-noise, single-bird recordings. However, audio from automated field recorders contains too much noise for such methods, so a more robust segmentation method is required. We propose a supervised timefrequency audio segmentation method using a Random Forest classifier, to extract syllables of bird call from a noisy signal. When applied to a test data set of 625 field-collected audio segments, our method isolates 93.6% of the acoustic energy of bird song with a false positive rate of 8.6%, outperforming energy thresholding.

Automatic classification of flying bird species using computer vision techniques

“…A number of existing attempts to automate the identification of birds have used audio rather than visual signals, such as (Briggs et al, 2009;Neal et al, 2011;Lopes et al, 2011;Bardeli et al, 2010). The use of audio signals has some attractive features; species typically have distinctive calls, and no line of sight is necessary to detect audio.…”

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

“…Las vocalizaciones de algunos animales se han jerarquizado, por ejemplo en las aves, éstas se dividen en cuatro niveles: notas, sílabas, frases y cantos [42]; en la literatura se encuentran varios trabajos orientados a la clasificación de sílabas [4], [32], [35], [37], [42], [43]. Los métodos de segmentación más básicos están diseñados para detectar las regiones con mayor energía [44], sin embargo, en otros mé-todos más elaborados se busca detectar trayectorias características en las representaciones tiempo-frecuencia [37], [41], [45].…”

Clusterized Mel Filter Cepstral Coefficients and Support Vector Machines for Bird Song Identification

“…computing a new feature vector on a window of n frames, to get new feature vectors that are representative of longer segments. The idea is close to the standard syllable extraction step that is used in most of methods for bird identification [12,2,1], but is much simpler to implement. In our case we considered segments of about 0.5 second duration (i.e.…”