Audrey: The First Speech Recognition System

Posted on October 13, 2014


downloadFrom mobile phones to personal computers and beyond, much of today’s technology utilizes speech recognition software—but where did speech recognition software get its start, and who pioneered the way for the advancements made today?

In the Beginning, There was Audrey

Before Dragon or Siri, there was Audrey, the “automatic digit recognizer.” The name suggested an ability to recognize numbers, but the technology of the time severely limited the system’s capacity to do so. Being a creation of 1952, Audrey could only distinguish between ten numbers (“0” through “9”). Despite her small vocabulary, Audrey was a marvel of science.

Oh, the Circuitry!

As with most first-generation electronics, Audrey was rather large. What can now be accessed from a smart phone the size of your hand required the walls of New York’s Bell Laboratories to house. Audrey’s relay rack alone stood 6 feet tall; one can only guess how much space the analog circuit of the system consumed with its amplifiers, integrators, and filters—and all for just ten numbers.

From Speech to Recognition

Creators K. H. Davis, R. Biddulph, and S. Balashek had their work cut out for them when they began their work on Audrey. Feeding speech sounds into a machine and receiving confirmation of hearing those sounds required special circuitry and complex processes. The procedure from input to output went something like this:

  1. The speaker recited the chosen digit or digits into an ordinary telephone, making sure to pause for 350 milliseconds between each word.
  2. Audrey listened to the speaker’s input, and, using the fundamentals of memory and matching (and a bunch of electrical operations I can’t even imagine), sorted the speech sounds into electrical classes that matched established reference patterns, which were drawn electronically beforehand and held within an analog memory.
  3. Audrey responded visibly by flashing the appropriate light.

Even with all her special gadgets, Audrey had to be fine-tuned to work at her optimal capacity. This involved adjusting the system to a particular speaker—meaning that Audrey’s use was far from universal. Not just anyone could use the system. To maintain the 97% accuracy of digit recognition boasted by Bell Laboratories, Audrey had to be familiar with a given speaker.

Audrey’s Legacy

Advances in computer technology have increased the efficiency and processing ability of speech recognition software, building upon Audrey’s 10 digits to encompass a seemingly innumerable amount of words. Today, untold numbers of people use speech recognition software to complete business duties, perform personal tasks, and even write their books.

As writers, especially, it is important to remind ourselves how far we have come since the start of our writing journey. Like Audrey, our first endeavors were limited by our mental capacity and skill, but we have made great advancements over time. Writing is a continuous struggle. It is easy to take our progress for granted when in that struggle, but we must remind ourselves when frustrated by our work or by ourselves that with strife, comes growth.

Audrey gave way to Dragon Dictate and Siri within a sixty-year period. How far have you come since you began writing?

Jennifer Warren, Writer and Editor, Asta Publications

Asta Publications has a long history of helping writers tell their stories and get published. Since 2004, Asta Publications has helped hundreds of authors bring their book concepts to life and we are ready to help you too! Our dedication to our authors is unmatched. We deliver first-class products and services that are accurate, high quality, and exceed our authors’ expectations.

Posted in: Publishing