Cemil Demir Email & Phone Number

Istanbul, Türkiye

Istanbul, Turkey

Design and implement speech assessment system for mobile devices.Implement content generation pipeline using LLMs.

Istanbul, Turkey

Design and implementation of automatic speech assessment system for mobile language learning application

Development of Wave-RNN Based Speech Synthesis Systems for Audiobook Generation, Automatic data generation for Speech Synthesis Model Training

Data augmentation system development for CNN Based Speech Enhancement Systems, Multi-Channel Speech Enhancement System with CNNs

Istanbul, Turkey

Development of Deep Learning based Speech Recognition, Speech Synthesis, Voice Conversion and Speaker Verification Systems.

Istanbul, Turkey

Deep LSTM based Text-to-Speech and Speech-to-Text systems for different languages.

Recently, we are working on improving our speech recognizer that can use RNN-BLSTM based acoustic models.We developed a deep neural network based speech recognition engine for Turkish Language. We trained DNN acoutic models using Kaldi. We implemented DNN based decoder in Java.We participated in NIST Keyword Search Evaluation (OpenKWS) 2015. We integrated RNN-BLSTM acoustic models to Kaldi system. We trained RNN-BLSTM acoustic models using CURRENNT system. We applied system combination technique and outperformed the best DNN based KWS system results.We developed a keyword search system for Turkish call-center telephone speech data.We developed a Kaldi based Turkish ASR system.

We designed and implemented our own ASR system.We developed a GMM based acoustic model trainer in Java.We developed a speech recognition engine which uses GMM acoustic models and n-gram language models in Java.We implemented sub-word language modeling approach for Turkish language. We applied stem-ending based language model inside our recognizer.We developed Optical Character Recognition (OCR) systems for Turkish and Ottoman handwritten data.We developed RNN-BLSTM based OCR system using RNNLIBWe developed a search algorithm based on RNNLIB output label probabilities.

I worked on various speech processing related projects at Speech and Language Processing Systems Laboratory. My main achievements are: - We developed a speech recognition system based on CMU Sphinx tools. - I developed a language identification system based on shifted-delta ceptsrum features and parallel phone recognition followed by language modeling (PPRLM) techniques. -I implemented a speaker identification system based on GMM.

Ankara, Turkey

Doctor Of Philosophy (Ph.D.), Electrical-Electronics Engineering, 3.81

In my phd study, I worked on single-channel source separation problem using machine learning techniques. I focused on background music.

Msc, Electrical-Electronics Engineering, 3.50

Bsc, Electrical-Electronics Engineering, 3.11