Imomi sensor: Artificial Intelligence & Sport

We started experimenting with a Sensor, an IMU sensor, and Artificial Intelligence software to enrich data experience in Tennis practices.

An IMU sensor (Inertial Measurement Unit) is an hardware device that is able register linear acceleration as also the gyroscopic acceleration (i.e. angular acceleration) along the three axes X, Y and Z (see fig. 1).
Fig. 1 — Example MPU 6050 IMU device

We created an HW integrated device Imomi, able to transmit via WiFi all IMU data to another device which could be a computer or an Android APP enabled with WiFi connectivity. The figure shows the prototypal sensor installed on a Tennis Racquets and transmitting data over WiFi. The sensor is able to presents an Access Point WiFi where client devices have to connect to receive data through WiFi.

Data Collection

As an example of data coming from the sensor, we have collected data through a Python script and plotted them using Pandas library. Pandas has an easy plotting capabilities to represent the time series of IMU data coming from the Imomi sensor.

In the figures below you can see how the repeated rotations done with the sensor, are represented in a Pandas plot.

Pandas Plot of IMU data

The magic of Artificial Intelligence starts here.

With the data collected throught the Python script, we trained a Neural Network, before we talked also of the availability due the WiFi connection to interact with an Android APP.

The next step forward was to design an android APP to recognize the rotation burst made with the sensor as soon they appear in the data frame flowing from the sensor to the APP.

The APP worked very well recognizing the rotation from the statis. Moreover the APP was able to identify the rotations burst distinguishing them by different repeated burst movements, as could be rapid oscillations of the sensor.

Below there is a screenshot of the APP numbering the Rotation recognized while it is connected to the Imomi sensor

This experiment was very interesting and demonstrate that using market low cost HW and all the Python and Android SW ecosystem to realize an intelligent system able to recognize (Sport) gesture and provide Sport feedback to player analyzing the recognized movement and also computing, further on, BioMetrics and BioMechanical parameters of these movements.

We are developing new frontier of Sport sensor and Artificial Intelligence sw to Action Sports, for more information or to collaborate on: hw, sw, AI, app, please write to: @