Cassio Batista

I have a BSc degree in Computer Engineering (2016) and a MSc degree in Computer Science (2017) conferred by Federal University of Pará (UFPA), Brazil. I’ve also spent one year (2014) as exchange student at the Óbuda University (OE) in Budapest, Hungary. Currently, I am a PhD student in Computer Science at UFPA, in Brazil. Some of my professional experiences include speech recognition for desktop, mobile and embedded platforms; as well as in developing alternative control interfaces for people with disabilities.


  • Speech Recognition
  • Natural Language Processing
  • Assistive Technology
  • Linux-based SBC Computers
  • Human-Computer Interaction


  • PhD student in Computer Science, 2020

    Current role at Federal University of Pará

  • MSc in Computer Science, 2017

    Federal University of Pará

  • BSc in Computer Engineering, 2016

    Federal University of Pará


What am I (supposed to be) good at?

Speech Recognition

CMU Sphinx, Kaldi, HTK, and Julius for Brazilian Portuguese

Embedded Hardware

Raspberry Pi, Beagleboard, C.H.I.P., Arduino, etc.

Assistive Technology

Software- and hardware-based solutions for people with visual, hearing and upper-limb motor disabilities

Genetic Algorithms

Routing and wavelength assignment (RWA) simulation in WDM-optical networks

Linux & Tools

Arch, XMonad, Vim, Git, Python, C, LaTeX, etc.

Speech Synthesis

HTS, Klatt



PhD in Computer Science

Federal University of Pará (UFPA)

Dec 2017 – Dec 2021 Belém, Brazil

Speech-based technologies:

  • Kaldi ASR for Brazilian Portuguese
  • Utterance copy TTS in English using Klatt and deep learning techniques

MSc in Computer Science

Federal University of Pará (UFPA)

Mar 2017 – Dec 2017 Belém, Brazil

A universal remote control system in C++ for people with upper-limb motor disabilities, so they could control a TV via alternative methods.

  • OpenCV for head gesture recognition
  • PocketSphinx for speech recognition
  • Adaptive switches in hardware

Research Internship


Mar 2016 – Dec 2016 Belém, Brazil
A simulator in Python for the routing and wavelength assignment (RWA) problem over transparent, wavelength-multiplexed optical networks using Genetic Algorithms.

Summer Internship

Óbuda University (OE)

Mar 2014 – Jan 2015 Budapest, Hungary

Development of speech (English) modules for controlling Teki: a personal home assistant, Turtlebot-based robot

  • PocketSphinx desktop on Linux + ROS (offline)
  • Android’s Google ASR (online Wi-Fi UDP connection)

Research Internship

Federal University of Pará (UFPA)

Jan 2012 – Feb 2016 Belém, Brazil

Development of resources and applications for spech recognition in Brazilian Portuguese:

  • PyQt4 CFG/BNF grammar tester for Julius
  • Acoustic model training on CMU Sphinx for KDE Simon Listens
  • Android client + Julius server vs. Google’s Android ASR

Recent Posts

Free Online Courses

I decided to post some comments about some excellent online courses related to computing and engineering that I’ve started to take during Covid-19 self-quarantine. Literally all of them are available for free on YouTube.


Head Remote

A system where user’s head gestures are translated into remote commands to electronic devices

Speech Remote

A remote control system that translates the user’s spoken words into commands to electronic devices


Speech recognition and TV remote control using Android and BeagleBone Black

Recent Publications

Towards a Free, Forced Phonetic Aligner for Brazilian Portuguese Using Kaldi Tools

Forced phonetic alignment in Brazilian Portuguese using Kaldi tools.

A Parallel Strategy for a Genetic Algorithm in Routing Wavelength Assignment Problem Using GPU with CUDA

Routing and wavelength assingment simulador on NVIDIA CUDA GPUs.

Evaluating Alternative Interfaces Based on Puff, Electromyography and Dwell Time for Mouse Clicking

Statistical comparison among three different types of mouse click: mouth puffing, EMG and dwell-time. Two out of these three methods have been developed in hardware and their schematics been open-sourced.

Utterance Copy in Formant-based Speech Synthesizers Using LSTM Neural Networks

Estimating the input parameter of Klatt88 formant-based speech synthesizer with long short-term memory neural nets (LSTM).

Baseline Acoustic Models for Brazilian Portuguese Using Kaldi Tools

The first attempt to create scripts and baseline acoustic models for Brazilian Portuguese using Kaldi tools.