The device follows on word
Embedded systems with voice control
In the 1970s, when Captain Kirk in the Starship Enterprise wanted to know something, he simply talked to the computer. It's fascinating that such a thing is no longer science fiction today. We have all long since become accustomed to sending search commands in our cell phones by voice command. Even in inexpensive vehicles, voice controls are now standard.
So embedded developers are also increasingly faced with the challenge of enhancing their devices by means of voice control.
What options are there for this?
Basically, a distinction is made between back-end and front-end systems. In the back-end system, the audio signal is only digitized on the device. Processing and evaluation of the speech are delayed by a service on a server somewhere on the Internet. This is how Siri, Google or Alexa work. The useful thing for the providers of these voice services is that they learn a lot about our wishes on the side, so that they can later sell targeted advertising. That's why these services are free for us. Applications with back-end systems are easy to implement. Instructions for this are available in many magazines of the Maker scene.
If an embedded system is not networked, if you need to respond to speech input without delays, or if you don't want to entrust speech data to a third-party service, it's better to rely on a front-end system. Speech processing, recognition and evaluation are then performed directly in the embedded system. If you have enough computing power in the device and rely on an operating system such as Embedded Linux, you can choose from a large number of free and commercial software components to implement voice control. Alternatively, voice control can also be integrated as a turnkey chip or module in one's own embedded system.
Front-end systems, unlike back-end systems, are limited in the scope of speech, since local resources must be sufficient for this purpose. In principle, a back-end system somewhere in the cloud sets no limits here.
One thing that must not be forgotten in voice control is the audio part of the embedded system. It is important to separate the actual speech from background noise and to digitize it as noise-free as possible. Here it has proven useful to use several microphones, which is supported by most speech recognition chips and modules. And a little analog know-how is still helpful to implement a speech control successfully.
Translated with www.DeepL.com/Translator (free version)