Reduction of Noise and Interfering Signals

Reduction of Noise and Interfering Signals for POCKET SPHINX Speech Recognition Software

In a modelhome scaled 1:25, actions will take place after special words or word combinations like light on/off, aircondition on/off, blinds/no blinds, are spoken into a microphone. The core of the construction is a state-of-the-art speech recognition software named “Pocket Sphinx” which is based on an FPGA (Field Programmable Gate Array), an electronic unit.

However, after the implementation and completion, it showed that the system didn't react because of background noises.

Together with the college of higher education (Fachhochschule) Hagenberg we sought a solution to this problem. The goal of this project was not only to reduce noises in the audio range but also to make dynamic disturbances ineffective so to feasibly achieve speech recognition with the smallest possible error rate.

Different noise reduction techniques were hereby brought to application, for example static and dynamic filtering with signal processors, PLL-regulation circuits as tracking filters, switched capacitor techniques, software solutions for embedded systems etc.

Interesting for Students

  • close teamwork with teachers and students from the college of higher education (Fachhochschule) and universities
  • up to date topic
  • future-oriented topic with continuous improvement in this special area
  • working with state-of-the-art working utilities
  • affordable tools
  • wide variety of appliance, for instance aid for physically challenged people.

Visits were arranged on both sides to keep track of development and presentations will also be held. Students hereby learned project management and project monitoring for external customers.


Through the co-operation of the college of higher education (Fachhochschule) Hagenberg and the school HTL Donaustadt and with the support of Carnegie Mellon University, Pennsylvania, USA, an effective system should be evolved which is capable of controlling a kind of house robot. It should be able to understand commands and execute them.

This project has been completed.