Utilizing human speech as user input is becoming more and more popular as current technology develops, and adding such capability to your products gives them the appearance of being up to date with modern standards. When integrating voice control functionality in devices, there are two main challenges that we must overcome: voice recognition accuracy and data security.
We will be demonstrating the design and build process of a system dedicated to integrating basic voice control functionality in any appliance by employing Syntiant’s TinyML dedicated hardware and an audio classification model built and optimized using the Edge Impulse platform.
The NDP101 Neural Decision Processor and an AMD21 Cortex-M0+ host processor are both included in the Syntiant TinyML board. It is a fantastic option for audio based applications granted its SPH0641LM4H microphone and the 5 GPIOs make it great for triggering an external output like a light or various other appliances. Combining it with a properly trained and optimized Edge Impulse model, it can be used to build systems that bring voice control functionality to devices that seldom make use of it.
Built for Edge Impulse by the Zalmotek team