On loss functions and CNNs for improved bioacoustic signal classification
Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based...
Ausführliche Beschreibung
Gespeichert in:
| Autor*in: |
Xie, Jie [verfasserIn] |
|---|
| Format: |
E-Artikel |
|---|---|
| Sprache: |
Englisch |
| Erschienen: |
2021transfer abstract |
|---|
| Übergeordnetes Werk: |
Enthalten in: GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY - 2014, an international journal on ecoinformatics and computational ecology, Amsterdam [u.a.] |
|---|---|
| Übergeordnetes Werk: |
volume:64 ; year:2021 ; pages:0 |
| Links: |
|---|
| DOI / URN: |
10.1016/j.ecoinf.2021.101331 |
|---|
| Katalog-ID: |
ELV055077706 |
|---|
| LEADER | 01000caa a22002652 4500 | ||
|---|---|---|---|
| 001 | ELV055077706 | ||
| 003 | DE-627 | ||
| 005 | 20230626041147.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 220105s2021 xx |||||o 00| ||eng c | ||
| 024 | 7 | |a 10.1016/j.ecoinf.2021.101331 |2 doi | |
| 028 | 5 | 2 | |a /cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica |
| 035 | |a (DE-627)ELV055077706 | ||
| 035 | |a (ELSEVIER)S1574-9541(21)00122-9 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 082 | 0 | 4 | |a 610 |q VZ |
| 082 | 0 | 4 | |a 610 |q VZ |
| 082 | 0 | 4 | |a 530 |q VZ |
| 084 | |a 52.56 |2 bkl | ||
| 100 | 1 | |a Xie, Jie |e verfasserin |4 aut | |
| 245 | 1 | 0 | |a On loss functions and CNNs for improved bioacoustic signal classification |
| 264 | 1 | |c 2021transfer abstract | |
| 336 | |a nicht spezifiziert |b zzz |2 rdacontent | ||
| 337 | |a nicht spezifiziert |b z |2 rdamedia | ||
| 338 | |a nicht spezifiziert |b zu |2 rdacarrier | ||
| 520 | |a Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. | ||
| 520 | |a Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. | ||
| 700 | 1 | |a Hu, Kai |4 oth | |
| 700 | 1 | |a Guo, Ya |4 oth | |
| 700 | 1 | |a Zhu, Qibin |4 oth | |
| 700 | 1 | |a Yu, Jinghu |4 oth | |
| 773 | 0 | 8 | |i Enthalten in |n Elsevier |t GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY |d 2014 |d an international journal on ecoinformatics and computational ecology |g Amsterdam [u.a.] |w (DE-627)ELV022626204 |
| 773 | 1 | 8 | |g volume:64 |g year:2021 |g pages:0 |
| 856 | 4 | 0 | |u https://doi.org/10.1016/j.ecoinf.2021.101331 |3 Volltext |
| 912 | |a GBV_USEFLAG_U | ||
| 912 | |a GBV_ELV | ||
| 912 | |a SYSFLAG_U | ||
| 936 | b | k | |a 52.56 |j Regenerative Energieformen |j alternative Energieformen |q VZ |
| 951 | |a AR | ||
| 952 | |d 64 |j 2021 |h 0 | ||
| author_variant |
j x jx |
|---|---|
| matchkey_str |
xiejiehukaiguoyazhuqibinyujinghu:2021----:nosucinadnsoipoebocutcin |
| hierarchy_sort_str |
2021transfer abstract |
| bklnumber |
52.56 |
| publishDate |
2021 |
| allfields |
10.1016/j.ecoinf.2021.101331 doi /cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica (DE-627)ELV055077706 (ELSEVIER)S1574-9541(21)00122-9 DE-627 ger DE-627 rakwb eng 610 VZ 610 VZ 530 VZ 52.56 bkl Xie, Jie verfasserin aut On loss functions and CNNs for improved bioacoustic signal classification 2021transfer abstract nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Hu, Kai oth Guo, Ya oth Zhu, Qibin oth Yu, Jinghu oth Enthalten in Elsevier GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY 2014 an international journal on ecoinformatics and computational ecology Amsterdam [u.a.] (DE-627)ELV022626204 volume:64 year:2021 pages:0 https://doi.org/10.1016/j.ecoinf.2021.101331 Volltext GBV_USEFLAG_U GBV_ELV SYSFLAG_U 52.56 Regenerative Energieformen alternative Energieformen VZ AR 64 2021 0 |
| spelling |
10.1016/j.ecoinf.2021.101331 doi /cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica (DE-627)ELV055077706 (ELSEVIER)S1574-9541(21)00122-9 DE-627 ger DE-627 rakwb eng 610 VZ 610 VZ 530 VZ 52.56 bkl Xie, Jie verfasserin aut On loss functions and CNNs for improved bioacoustic signal classification 2021transfer abstract nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Hu, Kai oth Guo, Ya oth Zhu, Qibin oth Yu, Jinghu oth Enthalten in Elsevier GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY 2014 an international journal on ecoinformatics and computational ecology Amsterdam [u.a.] (DE-627)ELV022626204 volume:64 year:2021 pages:0 https://doi.org/10.1016/j.ecoinf.2021.101331 Volltext GBV_USEFLAG_U GBV_ELV SYSFLAG_U 52.56 Regenerative Energieformen alternative Energieformen VZ AR 64 2021 0 |
| allfields_unstemmed |
10.1016/j.ecoinf.2021.101331 doi /cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica (DE-627)ELV055077706 (ELSEVIER)S1574-9541(21)00122-9 DE-627 ger DE-627 rakwb eng 610 VZ 610 VZ 530 VZ 52.56 bkl Xie, Jie verfasserin aut On loss functions and CNNs for improved bioacoustic signal classification 2021transfer abstract nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Hu, Kai oth Guo, Ya oth Zhu, Qibin oth Yu, Jinghu oth Enthalten in Elsevier GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY 2014 an international journal on ecoinformatics and computational ecology Amsterdam [u.a.] (DE-627)ELV022626204 volume:64 year:2021 pages:0 https://doi.org/10.1016/j.ecoinf.2021.101331 Volltext GBV_USEFLAG_U GBV_ELV SYSFLAG_U 52.56 Regenerative Energieformen alternative Energieformen VZ AR 64 2021 0 |
| allfieldsGer |
10.1016/j.ecoinf.2021.101331 doi /cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica (DE-627)ELV055077706 (ELSEVIER)S1574-9541(21)00122-9 DE-627 ger DE-627 rakwb eng 610 VZ 610 VZ 530 VZ 52.56 bkl Xie, Jie verfasserin aut On loss functions and CNNs for improved bioacoustic signal classification 2021transfer abstract nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Hu, Kai oth Guo, Ya oth Zhu, Qibin oth Yu, Jinghu oth Enthalten in Elsevier GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY 2014 an international journal on ecoinformatics and computational ecology Amsterdam [u.a.] (DE-627)ELV022626204 volume:64 year:2021 pages:0 https://doi.org/10.1016/j.ecoinf.2021.101331 Volltext GBV_USEFLAG_U GBV_ELV SYSFLAG_U 52.56 Regenerative Energieformen alternative Energieformen VZ AR 64 2021 0 |
| allfieldsSound |
10.1016/j.ecoinf.2021.101331 doi /cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica (DE-627)ELV055077706 (ELSEVIER)S1574-9541(21)00122-9 DE-627 ger DE-627 rakwb eng 610 VZ 610 VZ 530 VZ 52.56 bkl Xie, Jie verfasserin aut On loss functions and CNNs for improved bioacoustic signal classification 2021transfer abstract nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. Hu, Kai oth Guo, Ya oth Zhu, Qibin oth Yu, Jinghu oth Enthalten in Elsevier GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY 2014 an international journal on ecoinformatics and computational ecology Amsterdam [u.a.] (DE-627)ELV022626204 volume:64 year:2021 pages:0 https://doi.org/10.1016/j.ecoinf.2021.101331 Volltext GBV_USEFLAG_U GBV_ELV SYSFLAG_U 52.56 Regenerative Energieformen alternative Energieformen VZ AR 64 2021 0 |
| language |
English |
| source |
Enthalten in GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY Amsterdam [u.a.] volume:64 year:2021 pages:0 |
| sourceStr |
Enthalten in GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY Amsterdam [u.a.] volume:64 year:2021 pages:0 |
| format_phy_str_mv |
Article |
| bklname |
Regenerative Energieformen alternative Energieformen |
| institution |
findex.gbv.de |
| dewey-raw |
610 |
| isfreeaccess_bool |
false |
| container_title |
GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY |
| authorswithroles_txt_mv |
Xie, Jie @@aut@@ Hu, Kai @@oth@@ Guo, Ya @@oth@@ Zhu, Qibin @@oth@@ Yu, Jinghu @@oth@@ |
| publishDateDaySort_date |
2021-01-01T00:00:00Z |
| hierarchy_top_id |
ELV022626204 |
| dewey-sort |
3610 |
| id |
ELV055077706 |
| language_de |
englisch |
| fullrecord |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">ELV055077706</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230626041147.0</controlfield><controlfield tag="007">cr uuu---uuuuu</controlfield><controlfield tag="008">220105s2021 xx |||||o 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1016/j.ecoinf.2021.101331</subfield><subfield code="2">doi</subfield></datafield><datafield tag="028" ind1="5" ind2="2"><subfield code="a">/cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)ELV055077706</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(ELSEVIER)S1574-9541(21)00122-9</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">610</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">610</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">530</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="084" ind1=" " ind2=" "><subfield code="a">52.56</subfield><subfield code="2">bkl</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Xie, Jie</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">On loss functions and CNNs for improved bioacoustic signal classification</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2021transfer abstract</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zzz</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">z</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zu</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%.</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%.</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Hu, Kai</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Guo, Ya</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Zhu, Qibin</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Yu, Jinghu</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="n">Elsevier</subfield><subfield code="t">GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY</subfield><subfield code="d">2014</subfield><subfield code="d">an international journal on ecoinformatics and computational ecology</subfield><subfield code="g">Amsterdam [u.a.]</subfield><subfield code="w">(DE-627)ELV022626204</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:64</subfield><subfield code="g">year:2021</subfield><subfield code="g">pages:0</subfield></datafield><datafield tag="856" ind1="4" ind2="0"><subfield code="u">https://doi.org/10.1016/j.ecoinf.2021.101331</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_U</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ELV</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_U</subfield></datafield><datafield tag="936" ind1="b" ind2="k"><subfield code="a">52.56</subfield><subfield code="j">Regenerative Energieformen</subfield><subfield code="j">alternative Energieformen</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">64</subfield><subfield code="j">2021</subfield><subfield code="h">0</subfield></datafield></record></collection>
|
| author |
Xie, Jie |
| spellingShingle |
Xie, Jie ddc 610 ddc 530 bkl 52.56 On loss functions and CNNs for improved bioacoustic signal classification |
| authorStr |
Xie, Jie |
| ppnlink_with_tag_str_mv |
@@773@@(DE-627)ELV022626204 |
| format |
electronic Article |
| dewey-ones |
610 - Medicine & health 530 - Physics |
| delete_txt_mv |
keep |
| author_role |
aut |
| collection |
elsevier |
| remote_str |
true |
| illustrated |
Not Illustrated |
| topic_title |
610 VZ 530 VZ 52.56 bkl On loss functions and CNNs for improved bioacoustic signal classification |
| topic |
ddc 610 ddc 530 bkl 52.56 |
| topic_unstemmed |
ddc 610 ddc 530 bkl 52.56 |
| topic_browse |
ddc 610 ddc 530 bkl 52.56 |
| format_facet |
Elektronische Aufsätze Aufsätze Elektronische Ressource |
| format_main_str_mv |
Text Zeitschrift/Artikel |
| carriertype_str_mv |
zu |
| author2_variant |
k h kh y g yg q z qz j y jy |
| hierarchy_parent_title |
GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY |
| hierarchy_parent_id |
ELV022626204 |
| dewey-tens |
610 - Medicine & health 530 - Physics |
| hierarchy_top_title |
GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY |
| isfreeaccess_txt |
false |
| familylinks_str_mv |
(DE-627)ELV022626204 |
| title |
On loss functions and CNNs for improved bioacoustic signal classification |
| ctrlnum |
(DE-627)ELV055077706 (ELSEVIER)S1574-9541(21)00122-9 |
| title_full |
On loss functions and CNNs for improved bioacoustic signal classification |
| author_sort |
Xie, Jie |
| journal |
GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY |
| journalStr |
GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY |
| lang_code |
eng |
| isOA_bool |
false |
| dewey-hundreds |
600 - Technology 500 - Science |
| recordtype |
marc |
| publishDateSort |
2021 |
| contenttype_str_mv |
zzz |
| container_start_page |
0 |
| author_browse |
Xie, Jie |
| container_volume |
64 |
| class |
610 VZ 530 VZ 52.56 bkl |
| format_se |
Elektronische Aufsätze |
| author-letter |
Xie, Jie |
| doi_str_mv |
10.1016/j.ecoinf.2021.101331 |
| dewey-full |
610 530 |
| title_sort |
on loss functions and cnns for improved bioacoustic signal classification |
| title_auth |
On loss functions and CNNs for improved bioacoustic signal classification |
| abstract |
Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. |
| abstractGer |
Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. |
| abstract_unstemmed |
Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%. |
| collection_details |
GBV_USEFLAG_U GBV_ELV SYSFLAG_U |
| title_short |
On loss functions and CNNs for improved bioacoustic signal classification |
| url |
https://doi.org/10.1016/j.ecoinf.2021.101331 |
| remote_bool |
true |
| author2 |
Hu, Kai Guo, Ya Zhu, Qibin Yu, Jinghu |
| author2Str |
Hu, Kai Guo, Ya Zhu, Qibin Yu, Jinghu |
| ppnlink |
ELV022626204 |
| mediatype_str_mv |
z |
| isOA_txt |
false |
| hochschulschrift_bool |
false |
| author2_role |
oth oth oth oth |
| doi_str |
10.1016/j.ecoinf.2021.101331 |
| up_date |
2024-07-06T16:32:54.839Z |
| _version_ |
1803848069774573568 |
| fullrecord_marcxml |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">ELV055077706</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230626041147.0</controlfield><controlfield tag="007">cr uuu---uuuuu</controlfield><controlfield tag="008">220105s2021 xx |||||o 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1016/j.ecoinf.2021.101331</subfield><subfield code="2">doi</subfield></datafield><datafield tag="028" ind1="5" ind2="2"><subfield code="a">/cbs_pica/cbs_olc/import_discovery/elsevier/einzuspielen/GBV00000000001500.pica</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)ELV055077706</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(ELSEVIER)S1574-9541(21)00122-9</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">610</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">610</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">530</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="084" ind1=" " ind2=" "><subfield code="a">52.56</subfield><subfield code="2">bkl</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Xie, Jie</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">On loss functions and CNNs for improved bioacoustic signal classification</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2021transfer abstract</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zzz</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">z</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zu</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%.</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Frog population has been experiencing rapid decreases worldwide, which is regarded as one of the most critical threats to global biodiversity. Therefore, large volumes of frog recordings have been collected for assessing this decline. Most previous studies focused on the syllable segmentation based frog call classification, which is sensitive to the background noise. Our recent study has used 1D-CNN and cross-entropy loss for frog call classification. However, the use of 1D-CNN is sensitive to the background noise and the imbalance of the number of different frog species is not considered. Therefore, this study aims to investigate loss functions and imbalance learning for bioacoustic signal classification in continuous recordings. Specifically, four types of loss functions are compared including cross-entropy loss, weighted cross-entropy loss, focal loss, and twin loss, which are combined with three CNN architectures: 1D-CNN, 2D-CNN, and 1D-2D-CNN. In addition, random oversampling is used for improving the classification performance of two imbalanced datasets. Experimental results show that (1) 1D-2D-CNN model can achieve the performance for classifying both Australia and Brazil frog calls. (2) Focal loss is the best among four loss functions for classifying low SNR recordings. (3) The highest macro F1-score for classifying both Australia and Brazil frog recordings are 89.26%±0.36% and 93.47%±0.28%.</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Hu, Kai</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Guo, Ya</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Zhu, Qibin</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Yu, Jinghu</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="n">Elsevier</subfield><subfield code="t">GLOMERULAR FEATURES OF CARDIO-RENAL SYNDROME, A CASE CONTROLLED STUDY</subfield><subfield code="d">2014</subfield><subfield code="d">an international journal on ecoinformatics and computational ecology</subfield><subfield code="g">Amsterdam [u.a.]</subfield><subfield code="w">(DE-627)ELV022626204</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:64</subfield><subfield code="g">year:2021</subfield><subfield code="g">pages:0</subfield></datafield><datafield tag="856" ind1="4" ind2="0"><subfield code="u">https://doi.org/10.1016/j.ecoinf.2021.101331</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_U</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ELV</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_U</subfield></datafield><datafield tag="936" ind1="b" ind2="k"><subfield code="a">52.56</subfield><subfield code="j">Regenerative Energieformen</subfield><subfield code="j">alternative Energieformen</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">64</subfield><subfield code="j">2021</subfield><subfield code="h">0</subfield></datafield></record></collection>
|
| score |
7.39787 |