Data augmentation approaches for improving animal audio classification

In this paper we present ensembles of classifiers for automated animal audio classification, exploiting different data augmentation techniques for training Convolutional Neural Networks (CNNs). The specific animal audio classification problems are i) birds and ii) cat sounds, whose datasets are freely available. We train five different CNNs on the original datasets and on their versions augmented by four augmentation protocols, working on the raw audio signals or their representations as spectrograms. We compared our best approaches with the state of the art, showing that we obtain the best recognition rate on the same datasets, without ad hoc parameter optimization. Our study shows that different CNNs can be trained for the purpose of animal audio classification and that their fusion works better than the stand-alone classifiers. To the best of our knowledge this is the largest study on data augmentation for CNNs in animal audio classification audio datasets using the same set of classifiers and parameters. Our MATLAB code is available at https://github.com/LorisNanni.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: BioMediTech, Research group: Computational Biophysics and Imaging Group, Universita degli Studi di Padova, Italy
Contributors: Nanni, L., Maguolo, G., Paci, M.
Number of pages: 8
Publication date: 2020
Peer-reviewed: Yes

Publication information

Journal: Ecological Informatics
Volume: 57
Article number: 101084
ISSN (Print): 1574-9541
Original language: English
ASJC Scopus subject areas: Ecology, Evolution, Behavior and Systematics, Ecology, Modelling and Simulation, Ecological Modelling, Computer Science Applications, Computational Theory and Mathematics, Applied Mathematics
Keywords: Acoustic features, Animal audio, Audio classification, Data augmentation, Ensemble of classifiers, Pattern recognition
URLs: 
Source: Scopus
Source ID: 85082116223

Research output: Contribution to journalArticleScientificpeer-review

Classifying soil stoniness based on the excavator boom vibration data in mounding operations

The stoniness index of forest soil describes the stone content in the upper soil layer at depths of 20–30 centimeters. This index is not available in any existing map databases, and traditional measurements for the stoniness of the soil have always necessitated laborious soil-penetration methods. Knowledge of the stone content of a forest site could be of use in a variety of forestry operations. This paper presents a novel approach to obtaining automatic measurements of soil stoniness during an excavator-based mounding operation. The excavator was equipped with only a low-cost inertial measurement unit and a satellite navigation receiver. Using the data from these sensors and manually conducted soil stoniness measurements, supervised machine learning methods were utilized to build a model that is capable of predicting the stoniness class of a given mounding location. This study compares different classifiers and feature selection methods to find the most promising solution for this learning problem. The discussion includes a proposition for a meaningful measurement resolution of the soil’s stoniness, and a practical method for evaluating the variability of the stone content of the soil. The results indicate that it is possible to predict the soil stoniness class with 70% accuracy using only the inertial and location measurements.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Automation Technology and Mechanical Engineering, Metsäteho Oy
Contributors: Melander, L., Ritala, R., Strandström, M.
Publication date: 2019
Peer-reviewed: Yes

Publication information

Journal: Silva Fennica
Volume: 53
Issue number: 2
Article number: 10068
ISSN (Print): 0037-5330
Ratings: 
  • Scopus rating (2019): CiteScore 3.1 SJR 0.589 SNIP 1.007
Original language: English
ASJC Scopus subject areas: Forestry, Ecological Modelling
Keywords: Activity recognition, Spot mounding, Stoniness classification, Supervised machine learning
Electronic versions: 
Source: Scopus
Source ID: 85067380233

Research output: Contribution to journalArticleScientificpeer-review

Remediation of sedimented fiber originating from pulp and paper industry: Laboratory scale anaerobic reactor studies and ideas of scaling up

Anaerobic treatment of sedimented fibers collected from bottom of a bay that had been receiving pulp and paper mill wastewater for about 70 years were studied for the first time in semi-continuously fed continuously stirred tank reactors (CSTR). Anaerobic treatment of the fiber sediment was shown to be feasible, without dilution and with nitrogen and buffer supplement, at organic loading rates (OLR) up to 2.5 kg VS/m3d and hydraulic retention times (HRT) of 60 d resulting in methane yields of 201 ± 18 L CH4/kg VS. Co-digestion of sedimented fiber with sewage sludge at an OLR of 1.5 kg VS/m3d and HRT of 20 d resulted in a methane production of 246 ± 10 L CH4/kg VS. The techno-economic feasibility of mono and co-digestion process together with several case dependent factors such as maximum operable OLR, digestate utilization needs to be evaluated before making further conclusions for larger scale remediation applications.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Chemistry and Bioengineering, Research group: Bio- and Circular Economy, Finnish Consulting Group (FCG Suunnittelu ja tekniikka Oy)
Contributors: Chatterjee, P., Lahtinen, L., Kokko, M., Rintala, J.
Number of pages: 9
Pages: 209-217
Publication date: 15 Oct 2018
Peer-reviewed: Yes

Publication information

Journal: Water Research
Volume: 143
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2018): CiteScore 12.6 SJR 2.721 SNIP 2.486
Original language: English
ASJC Scopus subject areas: Ecological Modelling, Water Science and Technology, Waste Management and Disposal, Pollution
Keywords: Anaerobic digestion, Co-digestion, CSTR, Methane, Pulp and paper industry, Sedimented fiber

Bibliographical note

INT=keb, "Lahtinen, Leija"

Source: Scopus
Source ID: 85053165247

Research output: Contribution to journalArticleScientificpeer-review

Removal and recovery of uranium(VI) by waste digested activated sludge in fed-batch stirred tank reactor

This study demonstrated the removal and recovery of uranium(VI) in a fed-batch stirred tank reactor (STR) using waste digested activated sludge (WDAS). The batch adsorption experiments showed that WDAS can adsorb 200 (±9.0) mg of uranium(VI) per g of WDAS. The maximum adsorption of uranium(VI) was achieved even at an acidic initial pH of 2.7 which increased to a pH of 4.0 in the equilibrium state. Desorption of uranium(VI) from WDAS was successfully demonstrated from the release of more than 95% of uranium(VI) using both acidic (0.5 M HCl) and alkaline (1.0 M Na2CO3) eluents. Due to the fast kinetics of uranium(VI) adsorption onto WDAS, the fed-batch STR was successfully operated at a mixing time of 15 min. Twelve consecutive uranium(VI) adsorption steps with an average adsorption efficiency of 91.5% required only two desorption steps to elute more than 95% of uranium(VI) from WDAS. Uranium(VI) was shown to interact predominantly with the phosphoryl and carboxyl groups of the WDAS, as revealed by in situ infrared spectroscopy and time-resolved laser-induced fluorescence spectroscopy studies. This study provides a proof-of-concept of the use of fed-batch STR process based on WDAS for the removal and recovery of uranium(VI).

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Chemistry and Bioengineering, Helmholtz-Zentrum Dresden-Rossendorf, Ita-Suomen yliopisto
Contributors: Jain, R., Peräniemi, S., Jordan, N., Vogel, M., Weiss, S., Foerstendorf, H., Lakaniemi, A.
Number of pages: 9
Pages: 167-175
Publication date: 1 Oct 2018
Peer-reviewed: Yes

Publication information

Journal: Water Research
Volume: 142
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2018): CiteScore 12.6 SJR 2.721 SNIP 2.486
Original language: English
ASJC Scopus subject areas: Ecological Modelling, Water Science and Technology, Waste Management and Disposal, Pollution
Keywords: Adsorption, Desorption, Infrared spectroscopy, Sludge, STR, Uranium
Source: Scopus
Source ID: 85047810396

Research output: Contribution to journalArticleScientificpeer-review

Anaerobic digestion of 30−100-year-old boreal lake sedimented fibre from the pulp industry: Extrapolating methane production potential to a practical scale

Since the 1980s, the pulp and paper industry in Finland has resulted in the accumulation of fibres in lake sediments. One such site in Lake Näsijärvi contains approximately 1.5 million m3 sedimented fibres. In this study, the methane production potential of the sedimented fibres (on average 13% total solids (TS)) was determined in batch assays. Furthermore, the methane production from solid (on average 20% TS) and liquid fractions of sedimented fibres after solid-liquid separation was studied. The sedimented fibres resulted in fast methane production and high methane yields of 250 ± 80 L CH4/kg volatile solids (VS). The main part (ca. 90%) of the methane potential was obtained from the solid fraction of the sedimented fibres. In addition, the VS removal from the total and solid sedimented fibres was high, 61–65% and 63–78%, respectively. The liquid fraction also contained a large amount of organics (on average 8.8 g COD/L), treatment of which also has to be considered. The estimations of the methane production potentials in the case area showed potential up to 40 million m3 of methane from sedimented fibres.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Chemistry and Bioengineering
Contributors: Kokko, M., Koskue, V., Rintala, J.
Number of pages: 9
Pages: 218-226
Publication date: 15 Apr 2018
Peer-reviewed: Yes

Publication information

Journal: Water Research
Volume: 133
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2018): CiteScore 12.6 SJR 2.721 SNIP 2.486
Original language: English
ASJC Scopus subject areas: Ecological Modelling, Water Science and Technology, Waste Management and Disposal, Pollution
Keywords: Anaerobic digestion, Methane, Pulp and paper industry, Sedimented fibre
Source: Scopus
Source ID: 85041395267

Research output: Contribution to journalArticleScientificpeer-review

Biohydrogen production from xylose by fresh and digested activated sludge at 37, 55 and 70 °C

Two heat–treated inocula, fresh and digested activated sludge from the same municipal wastewater treatment plant, were compared for their H2 production via dark fermentation at mesophilic (37 °C), thermophilic (55 °C) and hyperthermophilic (70 °C) conditions using xylose as the substrate. At both 37 and 55 °C, the fresh activated sludge yielded more H2 than the digested sludge, whereas at 70 °C, neither of the inocula produced H2 effectively. A maximum yield of 1.85 mol H2 per mol of xylose consumed was obtained at 55 °C. H2 production was linked to acetate and butyrate production, and there was a linear correlation (R2 = 0.96) between the butyrate and H2 yield for the fresh activated sludge inoculum at 55 °C. Approximately 2.4 mol H2 per mol of butyrate produced were obtained against a theoretical maximum of 2.0, suggesting that H2 was produced via the acetate pathway prior to switching to the butyrate pathway due to the increased H2 partial pressure. Clostridia sp. were the prevalent species at both 37 and 55 °C, irrespectively of the inoculum type. Although the two inocula originated from the same plant, different thermophilic microorganisms were detected at 55 °C. Thermoanaerobacter sp., detected only in the fresh activated sludge cultures, may have contributed to the high H2 yield obtained with such an inoculum.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Chemistry and Bioengineering, Research group: Industrial Bioengineering and Applied Organic Chemistry, Institute for Water Education, Hydraulic and Environmental Engineering (IHE) Inst. for Water Education
Contributors: Dessì, P., Lakaniemi, A., Lens, P. N. L.
Number of pages: 10
Pages: 120-129
Publication date: 15 May 2017
Peer-reviewed: Yes

Publication information

Journal: Water Research
Volume: 115
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2017): CiteScore 11.5 SJR 2.601 SNIP 2.388
Original language: English
ASJC Scopus subject areas: Ecological Modelling, Water Science and Technology, Waste Management and Disposal, Pollution
Keywords: Biohydrogen, Butyrate, Dark fermentation, Inocula, Temperature, Xylose
Electronic versions: 
URLs: 
Source: Scopus
Source ID: 85014366720

Research output: Contribution to journalArticleScientificpeer-review

Effect of temperature on selenium removal from wastewater by UASB reactors

The effect of temperature on selenium (Se) removal by upflow anaerobic sludge blanket (UASB) reactors treating selenate and nitrate containing wastewater was investigated by comparing the performance of a thermophilic (55 °C) versus a mesophilic (30 °C) UASB reactor. When only selenate (50 μM) was fed to the UASB reactors (pH 7.3; hydraulic retention time 8 h) with excess electron donor (lactate at 1.38 mM corresponding to an organic loading rate of 0.5 g COD L-1 d-1), the thermophilic UASB reactor achieved a higher total Se removal efficiency (94.4 ± 2.4%) than the mesophilic UASB reactor (82.0 ± 3.8%). When 5000 μM nitrate was further added to the influent, total Se removal was again better under thermophilic (70.1 ± 6.6%) when compared to mesophilic (43.6 ± 8.8%) conditions. The higher total effluent Se concentration in the mesophilic UASB reactor was due to the higher concentrations of biogenic elemental Se nanoparticles (BioSeNPs). The shape of the BioSeNPs observed in both UASB reactors was different: nanospheres and nanorods, respectively, in the mesophilic and thermophilic UASB reactors. Microbial community analysis showed the presence of selenate respirers as well as denitrifying microorganisms.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Department of Chemistry and Bioengineering, Research group: Industrial Bioengineering and Applied Organic Chemistry, Institute for Water Education, UNESCO-IHE Institute for Water Education, Dept. of Civil-Environmental Engineering and Architecture, DICAAR, Université Paris-Est, UPEM, Department of Biochemical Engineering and Biotechnology, Indian Institute of Technology, Delhi, India, IRD UMR 206
Contributors: Dessì, P., Jain, R., Singh, S., Seder-Colomina, M., van Hullebusch, E. D., Rene, E. R., Ahammad, S. Z., Carucci, A., Lens, P. N. L.
Number of pages: 9
Pages: 146-154
Publication date: 1 May 2016
Peer-reviewed: Yes

Publication information

Journal: Water Research
Volume: 94
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2016): CiteScore 10.9 SJR 2.663 SNIP 2.583
Original language: English
ASJC Scopus subject areas: Water Science and Technology, Waste Management and Disposal, Pollution, Ecological Modelling
Keywords: Nitrate, Selenate, Selenium nanoparticles, Thermophilic, UASB
Source: Scopus
Source ID: 84959326965

Research output: Contribution to journalArticleScientificpeer-review

Data-based stochastic modeling of tree growth and structure formation

We introduce a general procedure to match a stochastic functional-structural tree model (here LIGNUM augmented with stochastic rules) with real tree structures depicted by quantitative structure models (QSMs) based on terrestrial laser scanning. The matching is done by iteratively finding the maximum correspondence between the measured tree structure and the stochastic choices of the algorithm. First, we analyze the match to synthetic data (generated by the model itself), where the target values of the parameters to be estimated are known in advance, and show that the algorithm converges properly. We then carry out the procedure on real data obtaining a realistic model. We thus conclude that the proposed stochastic structure model (SSM) approach is a viable solution for formulating realistic plant models based on data and accounting for the stochastic influences.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Department of Mathematics, Research group: MAT Inverse Problems
Contributors: Potapov, I., Järvenpää, M., Åkerblom, M., Raumonen, P., Kaasalainen, M.
Publication date: 2016
Peer-reviewed: Yes
Early online date: 3 Nov 2015

Publication information

Journal: Silva Fennica
Volume: 50
Issue number: 1
Article number: 1413
ISSN (Print): 0037-5330
Ratings: 
  • Scopus rating (2016): CiteScore 2 SJR 0.702 SNIP 1.116
Original language: English
ASJC Scopus subject areas: Ecological Modelling, Forestry
Keywords: Data fitting, Form diversity, Morphological plasticity, Plant model, Quantitative structure models, Stochastic functional-structural, Terrestrial lidar
Source: Scopus
Source ID: 84983200698

Research output: Contribution to journalArticleScientificpeer-review

Mesophilic anaerobic digestion of pulp and paper industry biosludge-long-term reactor performance and effects of thermal pretreatment

The pulp and paper industry wastewater treatment processes produce large volumes of biosludge. Limited anaerobic degradation of lignocellulose has hindered the utilization of biosludge, but the processing of biosludge using anaerobic digestion has recently regained interest. In this study, biosludge was used as a sole substrate in long-term (400 d) mesophilic laboratory reactor trials. Nine biosludge batches collected evenly over a period of one year from a pulp and paper industry wastewater treatment plant had different solid and nutrient (nitrogen, phosphorus, trace elements) characteristics. Nutrient characteristics may vary by a factor of 2-11, while biomethane potentials (BMPs) ranged from 89 to 102 NL CH<inf>4</inf> kg<sup>-1</sup> VS between batches. The BMPs were enhanced by 39-88% with thermal pretreatments at 105-134 °C. Despite varying biosludge properties, stable operation was achieved in reactor trials with a hydraulic retention time (HRT) of 14 d. Hydrolysis was the process limiting step, ceasing gas production when the HRT was shortened to 10 days. However, digestion with an HRT of 10 days was feasible after thermal pretreatment of the biosludge (20 min at 121 °C) due to enhanced hydrolysis. The methane yield was 78 NL CH<inf>4</inf> kg<sup>-1</sup> VS for untreated biosludge and was increased by 77% (138 NL CH<inf>4</inf> kg<sup>-1</sup> VS) after pretreatment.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Department of Chemistry and Bioengineering, Research group: Industrial Bioengineering and Applied Organic Chemistry, Urban circular bioeconomy (UrCirBio)
Contributors: Kinnunen, V., Ylä-Outinen, A., Rintala, J.
Number of pages: 7
Pages: 105-111
Publication date: 15 Dec 2015
Peer-reviewed: Yes
Early online date: 5 Sep 2015

Publication information

Journal: Water Research
Volume: 87
Article number: 11500
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2015): CiteScore 10.5 SJR 2.665 SNIP 2.49
Original language: English
ASJC Scopus subject areas: Water Science and Technology, Waste Management and Disposal, Pollution, Ecological Modelling
Keywords: Biogas, Hydrothermal pretreatment, Lignin, Methane production, Secondary sludge, Waste activated sludge

Bibliographical note

AUX=keb,"Ylä-Outinen, A."

Source: Scopus
Source ID: 84941946419

Research output: Contribution to journalArticleScientificpeer-review

Applicability of portable tools in assessing the bearing capacity of forest roads

Forest roads provide access to logging sites and enable transportation of timber from forest to mills. Efficient forest management and forest industry are impossible without a proper forest road network. The bearing capacity of forest roads varies significantly by weather conditions and seasons since they are generally made of poor materials and the constructed layers may be mixed with subgrade. A bearing capacity assessment is valuable information when trafficability is uncertain and rutting is obvious. In this study, bearing capacity measurements were carried out using the light falling weight deflectometer (LFWD), the dynamic cone penetrometer (DCP) and the conventional falling weight deflectometer (FWD). The aim was to compare their measurement results in relation to road characteristics and moisture conditions. Data were collected from 35 test road sections in four consecutive springs and during one summer. The test road sections had measurement points both on the wheel path and the centre line. The data show logical correlations between measured quantities, and the study presents reliable regression models between measuring devices. The results indicate that light portable tools, the DCP and the LFWD, can in most cases be used instead of the expensive falling weight deflectometer on forest roads.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Department of Civil Engineering, Research group: Earth Constructions, Life Cycle Effectiveness of the Built Environment (LCE@BE), Natural Resources Institute Finland (Luke), Natural Resources Institute Finland
Contributors: Kaakkurivaara, T., Vuorimies, N., Kolisoja, P., Uusitalo, J.
Number of pages: 26
Publication date: 2015
Peer-reviewed: Yes

Publication information

Journal: Silva Fennica
Volume: 49
Issue number: 2
Article number: 1239
ISSN (Print): 0037-5330
Ratings: 
  • Scopus rating (2015): CiteScore 2.2 SJR 0.64 SNIP 1.019
Original language: English
ASJC Scopus subject areas: Ecological Modelling, Forestry
Keywords: Dynamic cone penetrometer, Elastic modulus, Falling weight deflectometer, Light weight deflectometer, Stiffness
Source: Scopus
Source ID: 84929222878

Research output: Contribution to journalArticleScientificpeer-review

Decision-making in rights exporting: The integrated process

Rights exporting plays an essential role in the battle of fighting for Digital Rights Management (DRM) interoperability. The decision making process determines the results of rights exporting. In order to achieve optimal results in rights exporting, we leverage the process with algorithms for rights adaptation and rights decomposition. We also demonstrate how the proposed process can lead to improved results.

General information

Publication status: Published
MoE publication type: A4 Article in a conference publication
Organisations: Research Community on Data-to-Decision (D2D)
Contributors: Lu, W., Zhang, Z., Nummenmaa, J.
Number of pages: 8
Pages: 219-226
Publication date: 2013

Host publication information

Title of host publication: Proceedings of the 5th International Conference on Management of Emergent Digital EcoSystems, MEDES 2013
ISBN (Print): 9781450320047
ASJC Scopus subject areas: Information Systems and Management, Ecological Modelling
Keywords: DRM interoperability, rights exporting
Source: Scopus
Source ID: 84892648332

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Thermophilic anaerobic digestion of pulp and paper mill primary sludge and co-digestion of primary and secondary sludge

Anaerobic digestion of pulp and paper mill primary sludge and co-digestion of primary and secondary sludge were studied for the first time in semi-continuously fed continuously stirred tank reactors (CSTR) in thermophilic conditions. Additionally, in batch experiments, methane potentials of 210 and 230 m 3CH 4/t volatile solids (VS) added were obtained for primary, and 50 and 100 m 3CH 4/tVS added for secondary sludge at 35 °C and 55 °C, respectively. Anaerobic digestion of primary sludge was shown to be feasible with organic loading rates (OLR) of 1-1.4 kgVS/m 3d and hydraulic retention times (HRT) of 16-32 d resulting in methane yields of 190-240 m 3CH 4/tVS fed. Also the highest tested OLR of 2 kgVS/m 3d and the shortest HRT of 14-16 d could be feasible, if pH stability is confirmed. Co-digestion of primary and secondary sludge with an OLR of 1 kgVS/m 3d and HRTs of 25-31 d resulted in methane yields of 150-170 m 3CH 4/tVS fed. In the digestion processes, cellulose and hemicellulose degraded while lignin did not. pH adjustment and nitrogen deficiency needs to be considered when planning anaerobic digestion of pulp and paper mill wastewater sludges. © 2012 Elsevier Ltd.

General information

Publication status: Published
MoE publication type: A1 Journal article-refereed
Organisations: Tampere University of Technology, Department of Chemistry and Bioengineering, Urban circular bioeconomy (UrCirBio), Jyväskylän yliopisto, University of Jyväskylä
Contributors: Bayr, S., Rintala, J.
Number of pages: 8
Pages: 4713-4720
Publication date: 1 Oct 2012
Peer-reviewed: Yes

Publication information

Journal: Water Research
Volume: 46
Issue number: 15
ISSN (Print): 0043-1354
Ratings: 
  • Scopus rating (2012): CiteScore 8.1 SJR 2.914 SNIP 2.442
Original language: English
ASJC Scopus subject areas: Water Science and Technology, Waste Management and Disposal, Pollution, Ecological Modelling
Keywords: ADF, ADL, Anaerobic digestion, CSTR, FM, HRT, Methane, NDF, OLR, Primary sludge, Pulp and paper mill, SCOD, Secondary sludge, Thermophilic, TKN, TS, VFA, VS

Bibliographical note

Contribution: organisation=keb bio,FACT1=1

Source: researchoutputwizard
Source ID: 3899

Research output: Contribution to journalArticleScientificpeer-review