Research and Teaching
PhD Degree (to be announced … :D)
We offer the following research theme, please contact us for more detail and discussion (in Indonesian language is OK)
- Community-Based Signal Separation Evaluation Campaign. This is competition of sound separation as listed here, https://sisec.inria.fr/. The goal is very clear: create the best algorithm to separate sounds under different condition. The idea is to mimics human auditory system and most-recent algorithm like deep learning.
- Sound localization with Kinect
We already have kinect hardware. Inspired from here, we want to build listening device based on kinect. The goal is:
- Follow the azimuth (angle) of target sound source (first step)
- Follow the direction of target source (next step)
- Noise mapping based on Geographic Information System (GIS)
- Speech enhancement for underwater acoustic signal
- Deep learning source separation and speech recognition
- Automatic Fault type identification for vibration analysis
- Speech transcription
- Parametric Speaker for Selective Sound System
- Dodecahedron speaker for acoustic measurement
Currently, we use Surfer as tool to do noise map. Following the rapid growth of GIS, we want to adopt it by implementing noise map based on GIS. The tools we propose is ArcGIS with Spread-GIS or OrbisGIS Noise Map . For the initial step, we have database from the project located in East Java-based operation body.
After some experiment on signal enhancement by source separation method, we want to implement it for underwater acoustic signal. Please see our publication for reference.
In this theme, we would implement this Matlab DNN package for speech recognition by source separation method. Interest?
By using method developed in our lab, we could identify manually machinery fault by vibration and sound analysis. We would make it automatically as the vendor provided in the plant.
Automatic speech transcription is tool to convert speech voice to text. We starting this research theme based on our speech database in our VibrasticLab.
Some time, a speaker with selective direction is needed to keep privacy and maintain quietness. A parametric speaker can be designed to achieved this goal as show in the picture below.
|Parametric Speaker (Source: soundslazer.com)|
A dodecahedor speaker is usually priced expensive while it is not used regularly. We can built this speaker for our purpose.
|Contoh dodecahedron speaker di pasaran|
- Finite Element Analysis for Vibration and Acoustics
- Parallel programming for speech synthesize
What I want to do is to apply FEA based on Deal.II library to analyze Acoustic and Vibration problem. You may follow this github link to get some abstraction about crack propagation with finite element. Here is the example of membrane plot with paraview,
Speech synthesize needs long computation time to be generated by training from database by using learning methods such as Hidden Markov and machine learning. Currently, we only user serial programming to solve this problem. The proposed theme is to implement parallel programming for this task. I write an introduction to parallel programming for bash shell in Indonesian language here.
*) Tema – tema diatas bisa dijadikan judul/tema TA, PKM / Pimnas, atau kegiatan akademik saintifik lainnya. Do not hesitate to email me for discussion.
Semester Genap 2015/2016
- Fisika Bangunan (C)
- Pemrograman Komputer S1
- Pemrograman Komputer D3
- Menggambar Instrumentasi dan Sistem Perpipaan
Semester Ganjil 2015/2016
Semester Genap 2014/201
Semester Ganjil 2014/2015
- Optika S1
- Pengantar Teknologi Informasi & Komunikasi D3
- Sistem Pengukuran Geometri
- Fenomena Gelombang
- BioAustik (for Nursing Academy), 2009.
- Internet History, Security and Technology, Coursera, 2014.
- Fundamental of Design For Internet, Free Course. 2014.
- Pengenalan Sistem Operasi: Windows, Mac & Linux, free course for trainee, 2014.