Case Study

When designing a voice recognition system, it’s important that the speech system understands everyone speaking to it.

But what happens when a speech recognition system doesn’t account for children’s speech? Most speech recognition platforms are only designed with adults in mind. Child speech is very different than adult speech, and not all speech recognition devices are well-equipped to deal with the nuances.

From a technical standpoint, sounds generated by children average over 300Hz, compared to 210Hz in adult females, and 125Hz in adult males. Speech recognition devices that are trained to tune in to voices with lower frequencies will often miss much of what a child says.

Children often mispronounce words, too. The word ‘helicopter’ can be pronounced ‘hewwicopter,’ for example. This can cause chaos for speech recognition software that is trained to equate a set of pronunciations with a set of words.

When a leading global tech firm discovered that its speech recognition system couldn’t handle the nuances of children’s speech, they realized they needed to pull in a partner with specific expertise.

They turned to Appen for assistance.

Why did the global tech firm hire Appen to help with this speech collection project?

  • We have over 20 years’ experience collecting both spontaneous and scripted child speech
  • Our ability to recruit “minors,” thanks to our extensive network
  • Our skilled project managers with experience working with children to ensure optimal recordings

The global tech firm was able to expand into new markets, thanks to high-quality children speech data from Appen.

Read the case study to learn more.

Are you looking to improve the quality of your speech recognition systems? Click here to connect with one of our experts.