The MEMS market is one of the most rapidly growing sectors of the electronics industry.  In order to tempt the consumer to buy new products they have to be better or different from what they already have and incorporating sensors enables that.  A classic example is the inclusion of GPS, compass and motion sensors in the iPhone which made it stand apart from its rivals with a raft of exciting new features and applications that fired the consumers’ imaginations.  The game console market has similarly shown that motion sensing offers a whole new dimension of interaction.

So, as motion sensing and control is clearly going to be the enabling technology for a whole new generation of consumer devices, just how easy is it to implement in practice?

The good news

The market demand for MEMS is skyrocketing.  As it grows from a multimillion dollar industry to a multibillion dollar one, fabrication costs are coming down, performances are improving and power consumption is dropping.  

The bad news  

The increasing complexity of sensor configurations make the specialized task of signal processing even more challenging for OEM’s and systems integrators who often don’t have the necessary skill sets or intellectual property.  The tough demands of ever shortening Time To Market for consumer devices means that there isn’t the time to learn all the intricacies of sensor characterization and calibration along with data extraction and interpretation.    

Motion intelligence comes from a deep understanding of sensor systems, signal processing, biomechanics as well as a situational understanding of different application domains.  Movea has found that any given high level motion application can be composed of one or more of three basic types of features, those being:  detection, estimation and classification. 

• Detection is defined as determining whether an event occurred or not.

• Estimation is defined as measuring the motion properties of the event.

• Classification is defined as determining what the event was.

Let’s take a few examples.

In-air pointing 

To achieve in air pointing, a 2 axis gyroscope is employed in conjunction with an estimation algorithm that converts the rotational motion of the pointing device to a delta x / delta y translational motion of the cursor on the screen.  If required, a 3 axis accelerometer can be utilized to sense which way is down, compensating for any roll angle of the device ensuring that the cursor always move up when the device is rotated up.

Gesture recognition

These high level features are built from a combination of detection and classification.  The gesture recognition algorithm compares sensor data when a gesture is performed to the data in a gesture database, either choosing the best fitting gesture or rejecting it as an unrecognized gesture.  The algorithm can be built on top of a sensor set ranging from a simple 1 axis accelerometer up to a 9 axis sensor set utilizing accelerometers, gyroscopes and magnetometers.  The sensor set employed is highly dependent on size, nature and complexity of the gesture library.

Runner or walker applications 

Here one wants to estimate the trajectory of the foot.  In this application, we can take advantage of two very useful pieces of information: the user’s foot lands on the ground from time to time and therefore, if we can detect this, we know that at this time the sole of the shoe has a zero speed in the world’s frame of reference.  Moreover, we also know that the foot’s trajectory can be, with a good precision, considered as a trajectory in a plane, thus reducing the 6DOF.  Taking this into account, we can use fewer sensors to provide trajectory and use a 3A3M solution. These sensors provide a big advantage compared to gyro based systems as they are far less power hungry. The best trajectory plane is estimated and the trajectory in this plane is provided, so that any other parameters can be derived.

Think outside the box

As can be seen from these examples, different combinations of sensors can be used depending on the application.  9 axis sensor solutions can provide a wealth of information but the key data could be obtained by using fewer sensors, which saves costs and power consumption.  At Movea, we have found that the secret to success of adding e-Motion to the new generation of motion-enabled consumer devices is to cut costs by thinking outside the box as, by being clever, you can user fewer sensors than initially seem needed.

Bruno Flament is CTO of Movea SA and has 17 years experience in management at CEA-Léti dedicated to multi-sensor and autonomous micro-systems. He holds a PhD in signal processing and is the author of numerous technical documents and patents.

Read from EE TIMES