Presentation on theme: "„B I R D S M O N D” BIRD SOUND MONITORING DATABASE"— Presentation transcript:
1 „B I R D S M O N D” BIRD SOUND MONITORING DATABASE Robert Wielgat, Agnieszka Lisowska-Lis, Tomasz Potempa, Daniel KrólDepartment of Technology, State Higher Vocational School in Tarnów,Paweł KoziołComplex of Foothills Landscape Parks in TarnówZbigniew Bonczar, Damian Wiehle, Marcin LisDepartment of Zoology and Ecology, University of Agriculture in KrakówKazimierz WalaszInstitute of Environmental Sciences, Jagiellonian University in KrakówAntoni Ligęza, Tomasz ZielińskiAGH University of Science and Technology in Kraków
2 Reasons for avian monitoring Maintenance of biodiversity.Detection of ecological disasters.Monitoring actions in national and landscape parksMonitoring of environmental quality (Farmland Bird Index – FBI)Protection of endangered bird speciesProtection of plants in agricultureProtection of airports and planes
3 Problems of avian monitoring actions In order to monitor bird species there are organized monitoring programs involving large number of volunteers and skilled experts.In some actions necessary is catching wild birds (stressful).Collecting and analyzing results of observations are often difficult and troublesome. Mistakes are difficult to point out and correct.
4 Birdsmond Project Objectives The following task to accomplish are planned in the yearsBuilding recording device capable to record sounds and ultrasounds from several microphones and capable to measure GPS position.Building bird attracting deviceWriting a program for automatic recognition of bird voicesImplementing the data base accessed via Internet capable to store multimedial data coming from bird monitoring actions.Evaluating the system during monitoring actions and scientific expeditions.
5 General concept of the monitoring system Automatic ObserverBird voicerecognizerin unsupervisedmodeStationary digital recorderGuestMobile digital recorderExpert or AdministratorInformation systemGPS, movies, photos, weather informationStationary digital recorderMobile digital recorderObserverBird voice recognizer in supervised mode
6 Monitored AreaMonitored Area Includes Małopolska voivodeship in the south-east Poland especially terrain of Complex of Foothills Landscape Parks in the vicinity of Tarnów
7 Transects and observation points In order to experimentally evaluate all the system 6 transect and 7 observation points were chosen in Malopolska region, especially in Complex of Foothills Landscape Parks in Tarnów. Transects and observation points were defined, to represent variety of ecosystems: forests, parks, meadows, ponds, swamp, lakes, river bands, municipal terrains and refuse dump. More than 30 species vocalizations would be analysed during 3 years.
9 Transect: Tarnowiec-Radlna Satelite picture of Tarnowiec-Radlna transectTarnowiec-Radlna transect includes farmland, river banks and small village areas.
10 Transect: „Styr” Reservation „Styr” Reservation Transect includes mainly forest areas
11 Transect: „Stony Town” -Ciężkowice „Stony Town” Ciężkowice Transect includes forest and small rocks areas
12 Transect: „Krzyskie” - Ponds „Krzyskie” Ponds Transect includes ponds and swamp areas
13 Polichty Transect includes mainly forest areas and meadows Transect: PolichtyPolichty Transect includes mainly forest areas and meadows
14 Transect: Rożnowskie Lake Satelite picture of Rożnowskie Lake transect„Rożnowskie” Lake Transect includes artifical lake and forest areas
15 Observation Point: Zbylitowska Góra Zbylitowska Góra Observation Point is located on the Dunajec river bank near Collared Sand Martin (lat. Riparia riparia, Ital. topino) colony
16 Observation Point: PWSZ Tarnów Satelite picture of PWSZ Tarnów observation pointPWSZ Tarnów observation point is located in the municipal region
17 Observation Point: Mydlniki Mydlniki observation point is located in the experimental breeding farm of Peregrine Falcon (Lat. Falco peregrinus) (Ital. falco peregrino)
18 ConclusionThe general concept of the acoustical avian monitoring system has been presented.Preliminary Bird voice recognition experiments involving MFCC and HFCC features as well as DTW classification method have been carried out. Results of the experiments are promising to implement Bird Voice Recognizer.6 transects and 7 observation points have been determined along which initial test observations and recordings have been done.Preliminary web site version presenting encyclopaedic information on bird species has been prepared.
19 Future work and research Future research and work in years will include:Recognition experiments using TDSC, wavelet, spectral peaks features and HMM classification method.Implementation Bird Voice Recognizer as a computer program.Hardware implementation of digital recorder and bird attracting device.Implementation of data base and web site together with expert system.
20 Lanius collurio – red-backed shrike - L'averla piccola AcknowledgmentDescribed work is financed from grant of Polish Ministry of Science and Higher Education number N NLanius collurio – red-backed shrike - L'averla piccola
21 THANK YOU VERY MUCH FOR YOUR ATTENTION Emberizza citrinella – Yellowhammer- Lo zigolo gialloGarrulus glandarius – Jay- La ghiandaiaEmberizza citrinella – Yellowhammer- Lo zigolo gialloTHANK YOU VERY MUCH FOR YOUR ATTENTIONFor more information visit project website:
22 Stationary Digital Recorder LCDKeyboardWirelessTransceiverAntennaRealTimeClockMICROCONTROLLERFAT32BroadbandcondenserMicrophone x 4MemoryCardADCDACMicrophoneAmplifier x 4PoweramplifierAlluringspeakersReturn
23 Mobile Digital Recorder LCDKeyboardGPSAntennaRealTimeClockMICROCONTROLLERFAT32Broadbandcondensermicrophone x 4MemoryCardHeadphonesADCDACMicrophoneAmplifier x 4HeadphonesamplifierReturn
24 Bird Voice Recognizer – Unsupervised Mode Bird Voice Recognizer is a computer program capable to recognize bird species automatically using formerly recorded voice of the recognized bird species.Bird voice recognition is usually performed in the following stages:feature extractionclassificationBird voice recognition in unsupervised mode can be enhanced by an expert system using additional information like weather forecast, date and hour of the recordings, GPS position which are registered simultaneously with recognized bird voice.Return
25 Feature ExtractionThere are various features which can be extracted from bird voice signal for instance:TDSC (Time Domain Signal Coding) , spectral peaks, wavelets, MFCC (Mel Frequency Cepstral Coefficients), HFCC (Human Factor Cepstral Coefficients).Feature extraction in automatic bird voice recognition is sometimes preceded by initial signal processing like bandpass filtration, noise cancelation etc.So far MFCC and HFCC features were tested in the experiments obtainig promising ca. 92% recognition accuracy in the closed set experiment.Return
26 ClassificationThe most promising classification methods in bird voice recognition is Dynamic Time Warping (DTW) based word spotting and Hidden Markov Models (HMM) method.iYiXBIRD VOICE XBIRD VOICE Y1NM23451o1o2o3o4o5o6a23a22b2(o1)b2(o2)b2(o3)b4(o5)b3(o4)b4(o6)a34a45a12a33a44HMMDTWReturn
27 Bird Voice Recognizer – Supervised Mode In the supervised mode bird species recognized by Bird Voice Recognizer can be initially verified by the observer entering data to the system.The observer besides synchronized in time automatically captured information like bird voice, weather information, time and date, GPS position can also provide additional information like photos, movies and description of the bird or its behavior.Provided information together with initial observer verification can help in further bird species verification by an expert.Return
28 Information system Information system (IS) will consist of: Operating system (Linux distribution)Object-relational database (PostgreSQL);Data Warehouse supporting OLAP functions, homogeneously cooperating with database;Application server;IS will be bulit in three-tier architecture using MVC (Model – View – Controller) pattern.In order to store huge amount of multimedia data server will be equipped with almost 10 TB HDDs which are supervised by specialized RAID controller.
29 DatabaseServer with installed relational database will be a storage of data collected as a result of bird species recordings and observation. Database will contain various types of data especially audio files with bird voices, movies, photos, descriptions of particular bird which will be related to each other. Database will be divided into two parts: encyclopaedic one and experimental one.Return
30 Roles – Guest Guest will have authorization for: browsing encyclopaedic information about bird species;adding bird voices in order to recognize bird species automatically;Return
31 Roles – Observer Observer will have authorization for: adding results of observations;modyfing its own observations;all actions which are covered by guest role;ObserverReturn
32 Roles – Expert and Administrator Expert or AdministratorExpert will have authorization for:verifying information provided by observers or an automatic-observer;recognition questionable bird voices;Making statistical analysis supported by data mininggranting role of observer;all actions which are covered by observers role;Administrator will have authorization for:granting role of expert;all actions which are covered by an expert role;Return