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

Dufour, Olivier; Artières, Thierry; Glotin, Hervé; Giraudet, Pascale

doi:10.5772/56872

Cited by 18 publications

(9 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When using MFCCs, songs and calls are parameterized using descriptive measures derived from the temporal and spectral domains. This method has been used for automated recognition of calls of multiple avian species (Cai, Ee, Pham, Roe, & Zhang, 2007;Dufour, Artieres, Glotin, & Giraudet, 2014;Fagerlund 2007;Lee, Lee, & Huang, 2006;Potamitis, Ntalampiras, Jahn, & Riede, 2014). While MFCCs are a viable method for classifying bird songs, in certain situations they can be outperformed by other machine learning methods (Stowell & Plumbley, 2014).…”

Section: Alternative Processing and Statistical Techniquesmentioning

confidence: 99%

Quantitative acoustic differentiation of cryptic species illustrated with King and Clapper rails

Stiffler

Schroeder

Anderson

et al. 2018

Ecology and Evolution

View full text Add to dashboard Cite

Reliable species identification is vital for survey and monitoring programs. Recently, the development of digital technology for recording and analyzing vocalizations has assisted in acoustic surveying for cryptic, rare, or elusive species. However, the quantitative tools that exist for species differentiation are still being refined. Using vocalizations recorded in the course of ecological studies of a King Rail (Rallus elegans) and a Clapper Rail (Rallus crepitans) population, we assessed the accuracy and effectiveness of three parametric (logistic regression, discriminant function analysis, quadratic discriminant function analysis) and six nonparametric (support vector machine, CART, Random Forest, k‐nearest neighbor, weighted k‐nearest neighbor, and neural networks) statistical classification methods for differentiating these species by their kek mating call. We identified 480 kek notes of each species and quantitatively characterized them with five standardized acoustic parameters. Overall, nonparametric classification methods outperformed parametric classification methods for species differentiation (nonparametric tools were between 57% and 81% accurate, parametric tools were between 57% and 60% accurate). Of the nine classification methods, Random Forest was the most accurate and precise, resulting in 81.1% correct classification of kek notes to species. This suggests that the mating calls of these sister species are likely difficult for human observers to tell apart. However, it also implies that appropriate statistical tools may allow reasonable species‐level classification accuracy of recorded calls and provide an alternative to species classification where other capture‐ or genotype‐based survey techniques are not possible.

show abstract

Section: Alternative Processing and Statistical Techniquesmentioning

confidence: 99%

Quantitative acoustic differentiation of cryptic species illustrated with King and Clapper rails

Stiffler

Schroeder

Anderson

et al. 2018

Ecology and Evolution

View full text Add to dashboard Cite

show abstract

“…Multi-class classification can be done using SVM with a variety of different strategies; one of the more successful approaches is 'one-against-one' , where a binary classifier is trained for each pair of classes in the data-set (Hsu and Lin 2002). SVMs have been used to effectively classify bird songs (Cheng et al 2010;Dufour et al 2014), dolphin whistles (Esfahanian et al 2014), 88 different insect species (Noda et al 2016) and primates (chimpanzees, Fedurek et al 2016;marmosets, Turesson et al 2016).…”

Section: Introductionmentioning

confidence: 99%

Application of a semi-automated vocal fingerprinting approach to monitor Bornean gibbon females in an experimentally fragmented landscape in Sabah, Malaysia

2018

View full text Add to dashboard Cite

Vocal individuality has been documented in a variety of mammalian species and it has been proposed that this individuality can be used as a vocal fingerprint to monitor individuals. Here we provide and test a classification method using Mel-frequency cepstral coefficients (MFCCs) to extract features from Bornean gibbon female calls. Our method is semi-automated as it requires manual pre-processing to identify and extract calls from the original recordings. We compared two methods of MFCC feature extraction: (1) averaging across all time windows and (2) creating a standardized number of time windows for each call. We analysed 376 calls from 33 gibbon females and, using linear discriminant analysis, found that we were able to improve female identification accuracy from 95.7% with spectrogram features to 98.4% accuracy when averaging MFCCs across time windows, and 98.9% accuracy when using a standardized number of windows. We divided our data randomly into training and test data-sets, and tested the accuracy of support vector machine (SVM) predictions over 100 iterations. We found that we could predict female identity in the test data-set with a 98.8% accuracy. Using SVM on our entire data-set, we were able to predict female identity with 99.5% accuracy (validated by leave-one-out cross-validation). Lastly, we used the method presented here to classify four females recorded during three or more recording seasons using SVM with limited success. We provide evidence that MFCC feature extraction is effective for distinguishing between female Bornean gibbons, and make suggestions for future vocal fingerprinting applications.

show abstract

“…Most ML approaches in animal call classification take their lead from automated speech recognition by virtue of the commonalities between human speech and birdcalls. These ML approaches include supervised neural networks (including deep learning neural networks) [17]- [21], unsupervised neural networks [22], support vector machines [23]- [25], decision trees [26], [27], random forests [28], [29], and hidden markov model [30]- [34]. Despite the significant amount of research into the automated classification of birdcalls, there is not yet an adequate method for field recordings due to the challenges associated with birdcall classification, such as the high variability in calls.…”

Section: A Birdcalls In Acoustic Recordingmentioning

confidence: 99%

Data-Efficient Classification of Birdcall through Convolutional Neural Networks Transfer Learning

Efremova¹,

Sankupellay

Konovalov

2019

2019 Digital Image Computing: Techniques and Applications (DICTA)

View full text Add to dashboard Cite

Deep learning Convolutional Neural Network (CNN) models are powerful classification models but require a large amount of training data. In niche domains such as bird acoustics, it is expensive and difficult to obtain a large number of training samples. One method of classifying data with a limited number of training samples is to employ transfer learning. In this research, we evaluated the effectiveness of birdcall classification using transfer learning from a larger base dataset (2814 samples in 46 classes) to a smaller target dataset (351 samples in 10 classes) using the ResNet-50 CNN. We obtained 79% average validation accuracy on the target dataset in 5-fold cross-validation. The methodology of transfer learning from an ImageNet-trained CNN to a project-specific and a much smaller set of classes and images was extended to the domain of spectrogram images, where the base dataset effectively played the role of the ImageNet.

show abstract

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

Cited by 18 publications

References 3 publications

Quantitative acoustic differentiation of cryptic species illustrated with King and Clapper rails

Quantitative acoustic differentiation of cryptic species illustrated with King and Clapper rails

Application of a semi-automated vocal fingerprinting approach to monitor Bornean gibbon females in an experimentally fragmented landscape in Sabah, Malaysia

Data-Efficient Classification of Birdcall through Convolutional Neural Networks Transfer Learning

Contact Info

Product

Resources

About