A smart IoT platform requires three key building blocks: Connectivity, sensing and intelligence. On part 1 of ”DSP Enablers in the Internet of Smart Things” series, I addressed the connectivity aspect and explained how a DSP-based software PHY solution enables the creation of smart IoT devices through support for multiple connectivity standards. Let us now continue with a look at the sensing element.
There are multiple low-cost sensors in an Internet of Things (IoT) environment. Take the case of smart home, where an IoT home controller device has a microphone, can gather input from speech, process commands for speech recognition, carry out speaker verification, and in case of a suspicious noise, like breaking glass, can automatically call the security company.
Then, there are CMOS sensors in the house, which can perform tasks like motion detection, night vision and face recognition, and send messages to the owner about who is in the house.
So where does DSP fit in the sensing world? For a start, DSP enables you to analyze and aggregate the data from sensors. And there are more and more sensors generating a large amount of data and information, which needs to be processed at a very low power.
Internet of Sensors
There are multiple sensors for motion, sound, vision, health, and other environmental data. Biometric sensors are very important for wearable devices as they can interface with skin and muscles, and make health monitoring more durable. There are multiple motion sensing and multiple location and context awareness sensing apps using microphones, cameras and environmental sensors. Then, there are beacons and triangulation devices used to track location inside shopping malls and airports where GPS is not available.
These sensors may be very noisy, and there is a need for signal-cleaning — filtering, smoothing, calibrating, etc. — to extract the data. Inevitably, it leads to a lot of signal processing to calibrate and obtain meaningful data. And here comes the dilemma, as the mobile and IoT device OEMs demand that power consumption be few milliamps only for the always-on sensing apps.
Smart Sensing with DSP
In the early days, sensor fusion generally ran as part of an application processor software, but this approach consumed too much power. Next up, OEMs began using a sensor hub, which is usually a simple MCU doing simple accelerometer or motion sensing processing. Again, the problem is that OEMs require a very low power solution.
A DSP-based sensing solution allows taking down the power while carrying out multiple sensing tasks. A very low power DSP will listen all the time to voice commands via the microphone and wake up a main processor when needed. It can also execute intelligent context awareness through a learning process and pre-defined profiles. The DSP requirements further increase with the introduction of biometric sensors.
The CEVA-TL410 DSP core enables less than 150uW power consumption for always-on sensor fusion, voice trigger, face trigger and Bluetooth Low Energy. It can be embedded in a standalone sensor hub type chip device or it can be incorporated inside an audio codec chip. The DSP core and subsystem can also be integrated in application processor itself.
Making Sense of it all
It’s apparent that IoT devices need a lot of processing power to manage a plethora of sensors and they need to do it at a very low power. The DSP-based IoT solutions efficiently cater to signal-processing and low-power demands and thus can outperform systems built around CPU and MCU chip devices.
It’s also apparent that the connectivity and sensing worlds are intertwined in the IoT landscape. However, they need local intelligence to ensure that IoT devices act in a smart manner. The third and final blog of this series will show how local intelligence complements connectivity and sensing in the IoT environment.
You might also like
More from IoT
We’ve generally become comfortable with the idea that when you need digital signal processing - in the physical layers of …