Automatic Segmentation and Deep Learning of Bird Sounds

Koops, Hendrik Vincent; Balen, Jan Van; Wiering, Frans

doi:10.1007/978-3-319-24027-5_26

Cited by 15 publications

(10 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When ARU recording periods are limited to optimal call detection times near sunrise, it will be more cost‐effective to monitor Gambel's quail using ARUs for some studies. Advanced acoustic‐signal processing methods such as deep‐learning convolutional neural networks may be even more efficient at processing and successfully discriminating calling activity in large acoustic data sets further increasing the efficacy of nearly autonomous survey methods (Koops et al 2015).…”

Section: Discussionmentioning

confidence: 99%

Gambel's Quail Survey Variability and Implications for Survey Design in the Mojave Desert

Overton

Casazza

Connelly³

et al. 2020

Wildlife Society Bulletin

View full text Add to dashboard Cite

Careful design of a wildlife population monitoring strategy is necessary to obtain accurate and precise results whether the purpose of the survey is development of habitat suitability models, to estimate abundance, or assess site occupancy. Important characteristics to consider in survey design are sources of elevated variability, particularly within-subject variability, which increases the amount of data needed to achieve statistical certainty either in terms of population trend analysis, hypothesis testing, or statistical power. However, alternative objectives, such as associating counts with habitat characteristics, may benefit from increased variation among counts when differences covary with habitat measures. This difference can result in competing needs when developing survey protocols. We investigated the relative precision of differing gamebird monitoring protocols to identify methods with the greatest statistical efficiency. We assessed call-count transects using standard Breeding Bird Survey protocols (Passive call-counts) and modified by including longer survey periods and call playback (Active call-counts), autonomous recording units with supervised call detection (ARU-recorded calls), camera traps, and roadside covey-counts for Gambel's quail (Callipepla gambelii) in the Mojave Desert (CA, USA) during the spring of 2016. Active call-counts had the lowest within-site variation relative to estimated population index values, but Passive call-count transects may be more efficient for some purposes because more survey stations can be completed within a single survey timeframe. The ARU-recorded calls may provide a suitable alternative despite larger sample size needs, especially for occupancy surveys because multiple units can be deployed concurrently. The ultimate sample size required will depend on specific study objectives and scope of interest, but camera traps and breeding-season covey counts are not likely to meet objectives in desert environments.

show abstract

Section: Discussionmentioning

confidence: 99%

Gambel's Quail Survey Variability and Implications for Survey Design in the Mojave Desert

Overton

Casazza

Connelly³

et al. 2020

Wildlife Society Bulletin

View full text Add to dashboard Cite

show abstract

“…Manual segmentation of audio recordings is time consuming and tedious task. However, some approaches [5][6][7][8] have adopted manual segmentation while others have adopted automated segmentation of birdsong [9][10][11][12]. In general, detect any acoustic activity in audio and segment it as a syllable.…”

Section: Related Workmentioning

confidence: 99%

“…To achieve better results in segmentation, it is necessary to filter out this noise. Two types of high pass filters proposed in [11] are applied with some modifications. The first high pass filter is applied with a 500 Hz passband frequency instead of a 1000 Hz passband frequency, since the latter frequency caused the removal of the low-frequency sound of birds such as the Coppersmith Barbet.…”

Section: Segmentation Of Monosyllabic and Multisyllabic Birdsmentioning

confidence: 99%

“…For this reason, we have used two algorithms for segmentation: (i) extract the individual syllable present in the recording (referred to as Algorithm A); (ii) merge each consecutive syllable with the series of small-time intervals to extract the segment of syllables (referred to as Algorithm B). The algorithm is similar to the syllable extraction techniques described by [24] with some modifications [11]. The modification was made by adding preprocessing steps to eliminate the environmental noise and post-processing steps to cluster the syllables by applying Algorithm B.…”

Section: Segmentation Of Monosyllabic and Multisyllabic Birdsmentioning

confidence: 99%

See 1 more Smart Citation

Automatic Classification of Monosyllabic and Multisyllabic Birds Using PDHF

et al. 2021

View full text Add to dashboard Cite

Bioacoustics plays an important role in the conservation of bird species. Bio-acoustic surveys based on autonomous audio recording are both cost-effective and time-efficient. However, there are many bird species with different patterns of vocalization, and it is a challenging task to deal with them. Previous studies have revealed that many authors focus on the segmentation of bird audio without considering specific patterns of bird vocalization. Based on the existing literature, currently there is no work on the segmentation of monosyllabic and multisyllabic birds, separately. Therefore, this research addresses the aforementioned concern and also proposes a collection of audio features named ‘Perceptual, Descriptive, and Harmonic Features (PDHFs)’ that gives promising results in the classification of bird vocalization. Moreover, the classification results improved when monosyllabic and multisyllabic birds were classified separately. To analyze the performance of PDHFs, different classifiers were used in which Artificial neural network (ANN) outperformed other classifiers and demonstrated an accuracy of 98%.

show abstract

“…However, current implementations are: not generalizable to all species (Pearre, Perkins, Markowitz, & Gardner, 2017), very sensitive to hyperparameters (Ranjard & Ross, 2008), require pre-determined syllable classes and boundaries (Koumura & Okanoya, 2016) or struggle at recognizing subtle features that can be detected both by humans and birds (i.e. high false positives rate or low accuracy in test sets) (Fukuzawa, Marsland, Pawley, & Gilman, 2017;Koops et al, 2015).…”

Section: Fully Automatic Metric Extraction and Segmentationmentioning

confidence: 99%

Birdsong Phrase Verification and Classification Using Siamese Neural Networks

Rentería

Vallejo

Taylor

2021

Preprint

View full text Add to dashboard Cite

Bird vocalizations have been the focus of a wide variety of interdisciplinary studies in bioacoustics and neuroethology since they serve as models of motor control, learning and auditory perception. Yet, researchers have only begun to shed light on the structure and function of birdsong. Hypotheses abound, but still there is little agreement as how songs should be analyzed. One of the main challenges has been to classify acoustic units (syllables) from birdsong recordings, a task requiring robust classification algorithms capable of generalizing to unseen instances and dealing with data scarcity. Systematically detecting changes in syllable repertoires can help biologists to understand the origin and evolution of birdsong. The process of learning good features to discriminate among numerous and different sound classes is computationally expensive. Moreover, it might be impossible to achieve acceptable performance in cases where training data is scarce and classes are unbalanced. To address this issue, we propose a few-shot learning task in which an algorithm must make predictions given only a few instances of each class. We compared the performance of different Siamese Neural Networks at metric learning over the set of Cassini's Vireo syllables. Then, the network features were reused for the few-shot classification task. With this approach we overcame the limitations of data scarcity and class imbalance while achieving state-of-the-art performance.

show abstract

Automatic Segmentation and Deep Learning of Bird Sounds

Cited by 15 publications

References 9 publications

Gambel's Quail Survey Variability and Implications for Survey Design in the Mojave Desert

Gambel's Quail Survey Variability and Implications for Survey Design in the Mojave Desert

Automatic Classification of Monosyllabic and Multisyllabic Birds Using PDHF

Birdsong Phrase Verification and Classification Using Siamese Neural Networks

Contact Info

Product

Resources

About