1

Bird voice classification based on the combination

of feature extraction with the Convolutional Neural
Network(CNN)

SUPERVISED BY: PREPARED BY:

MST. AYESHA KHATUN HOSNA ARA SARKER ANIKA
SENIOR LECTURER ID: 19103219
DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING SALMA FARIHA AKTAR
ID: 19103212
 2
Contents
 Introduction

 Problem statement

 Motivation

 Research objective

 Literature Review

 Research methodology

 Result and Discussion

 Our contribution

 Limitations

 Future Work

 Conclusion

 References
 3
Introduction

 In this project we have detected the bird species using their voice.
 When a bird create a sound , it will detect this bird species.
 For classification method we used efficient network.
 4
Introduction(Cont.)

Problem Statement:

 Difficulties in identifying birds voice.

 Hard to recognize bird sound from long distance.
 5
Introduction(Cont.)

Research Motivation:

 Increasing hunting.
 Huge population.
 Sometimes many people don’t understand the sound of a bird and don’t recognize which
bird it is.
 6
Introduction(Cont.)

Research Objectives

 The main objective of this project is an automatic classification method to recognize bird
sound patterns.
 The method analyze the relationship between feature extraction method and bird signal
classification performance.
 7

Literature Review
 8
Bird Sound Recognition Using a
Convolutional Neural Network

 This paper authored by Agnes Incze , Henrietta-Bernadett Jancso , Zoltan Szil ´ agyi ´
† , Attila Farkas† and Csaba Sulyok (2018).
 In this paper, a CNN system classifying bird sounds is presented and tested through
different configurations and hyper parameters.
 9
Cont.
 10
Cont.

 The limitation of this paper is that the presented system is viable only for a low number
of classes.
 Their accuracy is 73%.
 11
Classification of bird sounds as an early warning method of forest fires using Convolutional
Neural Network (CNN) algorithm

 Silvester Dian Handy Permana , Gusti Saputra , Budi Arifitama , Yaddarabullah ,

Wahyu Caesarendra , Robbi Rahim (2021)
 This paper presents bird sounds classification study using Convolutional Neural Network
(CNN) method.
 The CNN method is used to classify bird sounds in the two conditions: (1) under normal
circumstances or conditions, and (2) under threated or panic condition.
 12
Cont.
 13
Cont.

 The anticipated value does not receive a particular mark from the system in comparison to
the available label options. It is intended that these flaws will be addressed such that the
system can explain both accurate and incorrect findings as well as results that cannot be
predicted or are parallel.
 Their accuracy is 78.5%.
 14
Birds Sound Classification Based on Machine
Learning Algorithms

 This paper authored by Aska E. Mehyadin , Adnan Mohsin Abdulazeez , Dathar Abas
Hasan and Jwan N. Saeed (2021) . Author AMA designed the study, performed the
statistical analysis. Author AEM wrote the protocol and wrote the first draft of the
manuscript. Author DAH managed the analyses of the study. Author JNS managed the
literature searches.
 In this paper they have used Mel-frequency cepstral coefficient (MFCC) and tested
through different algorithms to classify bird species.
 15
Cont.
 16
Cont.

 The limitation of this paper is difficult to hear the bird calls until the signal is filtered and
boost the volume.
 Their success rate is 78.40% and elapsed time are (39.4 seconds).
 17
Research Methodology

 We have used python language.

 We have used Mel Frequency Cepstral Coefficients (MFCC) methods for feature
extraction.
 We also use Efficient Net method that is an architecture of CNN, for predicting the bird
sound.
 18
Research Methodology
 19
Research Methodology
 20
Result & Discussion

Simulation Settings:

 Intermediate Language: Python

 Environment : Google Colab
 Dataset Collection : Kaggle.com
 21
Result & Discussion(Cont.)

 First of all we will pre-process the voice.

 22
Result & Discussion(Cont.)

LFPS MSS MSS with high pass filter

 23
Result & Discussion(Cont.)

 We have achieved accuracy 83%

 24
Result & Discussion(Cont.)

Table1: Comparison Table

Work Algorithm Accuracy

Convolutional Neural 73%

Network
Existing Work
K-Nearest Neighbor 78.11%

Naïve bayes 47.45%

MLP 74.68%

Our Work Efficient net 83%

 25
Our Contribution

 An efficient and accurate bird sound prediction system will be provided.

 We have used Efficient Net architecture for the higher accuracy.
 We predict bird more accurately.
 26
Limitation

 Resource Limitation
 Real Case Training Data is time-consuming and costly.
 27
Future Work

 Increasing the accuracy by using more data.

 Noise reduction and filtering to eliminate extremely high frequencies that are absent from
bird sounds are recommended for the preprocessing step.
 28
Conclusion

This study may complete the identification stage by reducing the amount of records from the
extraction results and employing constrained computational resources. We have made our
work like this- when we enter the audio of any bird that sound will first of all appeared with
its wave and also Linear-frequency power spectrogram, Mel-scaled spectrogram, Mel-scaled
spectrogram with high pass filter- 10 seconds. From the audio signal the bird will be detected.
 29
Reference

 Pulung N, Guruh F, Dwi p, Dewi P(2021), “Bird Voice Classification Based on Combination Feature
Extraction and Reduction Dimension with the K-Nearest Neighbor” International journal of Intelligent
Engineering & System.
 Agnes I, Henrietta B, Zoltan S, Attila F, Csaba S(2018), “Bird Sound Recognition Using a Convolutional
Neural Network” IEEE 16th International Symposium on Intelligent Systems and Informatics
 Aska E, Adnan M, Dathar A, Jwan N(2021) “Birds Sound Classification Based on Machine Learning
Algorithms” Asian Journal of Research in Computer Science
 Marcos V, kumar S, Pratek A, Nitin D, Szilard B(2021) “Weakly-Supervised Classification and
Detection of Bird Sounds in the Wild A BirdCLEF 2021 Solution” CLEF 2021 – Conference and Labs
of the Evaluation Forum
 Silvester D, Gusti S, Budi A, Wahyu C, Robbi R(2021) “Classification of bird sounds as an early
warning method of forest fires using Convolutional Neural Network (CNN) algorithm” Journal of King
Saud University – Computer and Information Sciences.
30
31