In 1952, at Bell Labs, the engineers Davis, Biddulph, and Balashek built the Automatic Digit Recognizer (Audrey), a rudimentary voice recognition system. Audrey was limited by the technology of the time but was able to recognize the numbers 0 to 9. The Audrey system, which processed the 10 digits through voice recognition, was 6 feet tall and covered the walls of Bell Labs, containing large numbers of analog circuits with capacitors, amplifiers, and filters. Audrey did the following three things:
- The Audrey system took the user's voice as input and put the voice into the machine's memory. The voice input was classified and pattern matching was performed against the predefined classes of voices for the numbers 0 to 9. Finally, the identified number was stored in memory.
- It flashed a light that represented the matching number.
- It was also able to communicate selected digits over the telephone.
Audrey performed what's known today as NLP, using ML with AI.
Although Audrey recognized voice input with an accuracy of 97% to 99%, it was very expensive and large in size, and it was extremely difficult to maintain its complex electronics. Thus, Audrey could not be commercialized. However, since the inception of Audrey, voice technology and research has continued to leap forward.