Network effects on the functionality of voice-controlled systems and applications

Abstract

The use of voice to control devices and applications provides a flexible, convenient and secure method for operating these systems. Users can control these systems using voice commands while they are in the proximity of the system to be controlled. Attempting to apply such control over the internet may be affected by several network impairments, e.g. jitter, errors, losses, etc. The effects of these impairments, while may not be felt by the normal human ear, can impact the operation of the software that processes these commands. This processing system is usually based on speaker identification and speech recognition components. In this paper, we study these network effects on the correct operation of such systems. We experiment with different types of network impairments and measure the effects on the voice-controlled operation. Our experiments show that network impairments affect the processing of voice commands to varying degrees. They also provide a comparative view between the behavior of command processing modules and the perceived quality of voice.