Voice recognition

The business goal

Our client is an entrepreneur with Electrical Engineering experience. His goal was to create an ebook reader that could be used by elderly people without too much effort. Users can use/control these readers by voice. The most important feature of the system is voice recognition that is used for handling ebooks, executing commands, navigating libraries and searching books. Bringing his hardware expertise, he needed an expertise in software development and Data Science research.

From the beginning, there were two main directions for the project. The first was to add online voice recognition due to the rising interest in that field and ability to connect to high-speed networks. The second was to improve offline recognition.

Technical side

For the first part we researched and tried commercially available solutions from Google, Microsoft and other providers. The second part had to do with using PocketSphinx library. It was used as a grammar based engine with strict rules and good quality of voice recognition in different situations (accent, noise, low volume and so on). Calibre for its part is used for ebook management.

On the technical side we used C-based Calibre software and PocketSphinx library. We also integrated Python-based APIs from online recognition services.


As a result we delivered a Proof-of-Concept system that could be readily used on an ebook device.

Simply a pleasure to work with Mediterra. Their team was both creative and methodical. they held regular meetings and followed up on any issues that arose. Code testing was integral to the process and is much appreciated.
Get in touch


    Contact us on Upwork

    Hire us

    Contact us on Linkedin

    Reach out us

    Moscow, 117335,