(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Developing a pre-processing algorithm of animal sound recognition on altera de2

Developing a pre-processing algorithm of animal sound recognition on altera de2 /Teh Yi Kai
Pada dasarnya, satu algoritma adalah kaedah langkah demi langkah atau satu set laluan untuk menyelesaikan masaslah. Untuk pengecaman bunyi haiwan, peringkat pra-pemprosesan adalah langkah yang sangat penting untuk mendapatkan maklumat yang sesuai terutama pada pengekstrakan cirinya. Isyarat bunyi biasanya dalam bentuk analog, ia terpaksa disalurkan melalui sistem pra-pemprosesan dan mengubah sifatnya ke dalam bentuk yang sesuai. Oleh itu, pra-pemprosesan algoritma yang sesuai mesti dibangunkan untuk menjana perkali yang perlu untuk mengkelaskan isyarat tertentu. Dalam projek ini, langkah demi langkah untuk manjana perkali untuk bunyi haiwan dibentangkan. Mel Frekuensi Cepstral Pekali (MFCC) digunakan untuk proses penyarian sifat. Bagi menjana pekali MFCC, Fast Fourier Transform (FFT), Mel Penapis Bank diikuti oleh jelmaan Diskret Kosinus (DCT) telah digunakan. Algoritma untuk pra-pemprosesan bunyi haiwan telah dibina dengan menggunakan kod verilog HDL dan kod C++. Papan litar Altera’s FPGA DE2 yang mengandungi pemproses asas Nios II digunakan untuk membina sistem tersebut. Selain itu, perisian seperti Altera Quartus II, Pembina SOPC dan Nios II EDS juga digunakan untuk menulis algoritma. Keputusan 12-titik MFCC yang diperolehi menunjukkan sistem pra-pemprosesan yang dibangunkan boleh fungsi seperti yang diharapkan. _______________________________________________________________________________________________________ Basically, an algorithm is a step-by-step method or a set of route for solving a problem. For animal sound recognition, pre-processing stage is the crucial step for getting appropriate information for feature extraction. Sound signal normally is analog form in nature, it had to pass through a proper pre-processing system and convert it into the suitable form. Hence, an appropriate pre-processing algorithm must be developed in order to generate coefficients for classify the input signal. In this project, a step by step to generate the coefficients for an animal sound is presented. Mel Frequency Cepstral Coefficient (MFCC) is used for the process of feature extraction. In order to generate the MFCC coefficients, Fast Fourier Transform (FFT), Mel Filter Bank followed by Discrete Cosine Transform (DCT) are used. The pre-processing algorithm of animal sound is built by using the verilog HDL code and the C++ code. Altera’s FPGA DE2 board which contains a Nios II soft core processor is used to build the system. Besides that, the software such as Altera’s Quartus II, SOPC Builder and Nios II EDS are used to write the algorithm. The obtained results of 12-point MFCC show that the developed Pre-processing system can be worked as expected.
Contributor(s):
Teh Yi Kai - Author
Primary Item Type:
Final Year Project
Identifiers:
Accession Number : 875004743
Language:
English
Subject Keywords:
algorithm; animal; extraction
First presented to the public:
6/1/2012
Original Publication Date:
7/2/2020
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 80
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2020-07-02 15:22:51.032
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Developing a pre-processing algorithm of animal sound recognition on altera de21 2020-07-02 15:22:51.032