Voice recognition system for speak!easy

speak!easy is a spatial interactive installation created for a festival of cocktail robotics Roboexotica 2021. It is designed as a speakeasy in the time of the prohibition. The authoritarian figure is played by a voice recognition system. To receive the cocktail, the visitor needs to communicate a password to the human on the other side, without being recognized by the computer.

5.2_Interaction.jpeg

The installation explores voice recognition systems. On one side it exposes their potential for mass surveillance, on the other, it tries to encourage visitors to find creative ways of bypassing it. With this new interaction possibility, users are encouraged to use methods of ‘speaking’ that are only understood by another person and are intentionally not understood by computers.

Presenting this installation in a public exhibition setting, allowed for an extensive field study where we collected all the successful interaction methods of bypassing a voice recognition system. These methods also provided valuable insights on what to watch out for when designing a real-world voice recognition application.

The most common ways of disguise were

  • singing

  • weird accents

  • mumbling

  • spelling each letter

  • whispering

Out of the most creative approaches was the use of Google translate to a language the system wouldn’t recognize, and the use of noise that made it impossible for the system to discern words.

Previous
Previous

Smart textile interfaces

Next
Next

Sound of Textile