For the past few blogs I have discussed about the voice recognition system that I have developed and implemented with the use of the Arduino Nano 33 Ble Sense.
Now, the Arduino 33 Ble Sense has a footprint of 45x18mm, and luckily for me, the company Seeed has released a similar device, based on the same chip nRF52840, at the end of December 2021, so 4 months ago from the point of this blog. And the dimensions are way lower, to 21x17.5, that is almost 60% less volume, which is a big benefit for my device, as I am trying to have everything fit on the wrist. Furthemore, the Seeed is also cheaper, so it would help with mass-manufacturing in terms of pricing, if the device would be for sale.
Here is the official website with details about the device: https://www.seeedstudio.com/Seeed-XIAO-BLE-Sense-nRF52840-p-5253.html
One important thing to note, is that the Seeed Xiao Sense only has Blueetooth, Battery Charging Chip, IMU and a MIC, it does not have any other sensors, especially environmental.
So the Arduino Nano 33 Ble Sense that was supposed to be used for sensor fusion, will be removed, together with the fusion functionality. In the end this is a positive change, as the sensors on the Nano were quite poor in terms of quality, and even when using Linear Statistical regression to calibrate them, the sensor would still have moments of poor outputs.
Below is a size comparison:
Because the hardware is exactly the same as the Nano 33 Ble Sense, the best trained CNN model for the speech recognition can just be compiled and uploaded, all the same.
Below is a video of the Xiao in action, classifing the voice commands:
And voila, it works with the exact same accuracy, while being considerably smaller, consuming less power, and not requiring a voltage regulator bypass to be usable at 3.3 volts, as the Nano 33 Ble Sense needs as seen here: https://support.arduino.cc/hc/en-us/articles/360014735580-How-to-bypass-the-DC-regulator-in-the-NANO-33-BLE-SENSE