Peak Detection
There is clear rhythmicity in the data, especially in the volleys and the groundstrokes. It would also be nice to zoom in on the volleys and see if we can see a clear signal here, even though these weren't picked up by the Zepp sensor.
The Sensor Logger app has an excellent accompanying website.
Included in this is peak detector software. Let's see how it does...
Impressive! But, it does look like it over sensed. However, if we set a threshold value, it ignores anything below it.
This lowers the detected peaks from 145 to 65. If you look closely, the peak detections aren't always were you think they would be. I think that this is due to the complicated nature of some of these waveforms. They aren't smooth up and down peaks in every case. I think that this already illustrates why all tennis tracking devices seem to use multiple sensors.
Let's see how the quaternionX signals does with the peaks detector. This signal was observed to be smoother than the raw gyro or accel signals.
This threshold level (which as far as I understand needs to manually be set) gives a stroke count of 43 volleys, and does indicate an alternating pattern, which might be indicative of forehand and backhand alternating. Since volleys weren't recorded by the other sensors, this is an exercise in unsupervised learning. But, this seems reasonable.
One other function included on the Sensor Logger website is a Fast Fourier Transform. For those, not in the know, a Fourier transform is a mathematical technique that takes a visualization from the time domain to the frequency domain. This means that the x-axis is now measuring frequency instead of time. It's a good way to measure repeatable signals.
Here's an FFT of the gyro signal:
This is probably not a highly elucidating visual, but it does seem to have peaks at around 0.5 HZ and 0.6, which would indicate about 2 seconds/shot. This seems to be about a little low compared to the quaternion reading of 43 volleys in 55 seconds but is perhaps because it's looking for a full cycle between shots, which might be two forehands or two backhands.
Volleys are the least interesting shot to measure though, so I bet it was a design decision to ignore volleys for the Zepp sensor, rather than a hardware limitation.
Let's look at some groundstrokes now....
Groundstrokes
Let's focus on the 6 powerful groundstrokes that were hit at the end of the session. Forehands and backhands altternating.
The forehands have clearly more powerful signals, although, interestingly the ZIQ measurement rates the backhands more highly. This is pretty difficult to understand, but also seems to repeat during the rest of the groundstroke sessions. It might be that the ZIQ system that we've designed somehow builds in bias towards backhand shots. This might be worth looking at another time, but doesn't matter for now.
Above is the peak detector of the accelerometerX. The rotation rate is also interesting, but the peak detector doesn't work as well here, so is left off. See below:
This is the first sensor signal that we've seen that has a clear directional aspect between forehands and backhands. This time of signal would clearly be used when deciding whether a stroke was a forehand or a backhand.
Fin
It's hard to know where we are, but we've certainly covered a lot of ground. We've seen that the Zepp Universal Sensor has a 3 axis accelerometer and a 2 axis gyro. The Babolat POP and the Zepp2 sensor probably have similar hardware, but we cannot access the data. All the sensors estimate racquet speed and spin, and either directly, or indirectly, measure sweet spot contact.
Both Zepp sensors can provide immediate shot feedback, along with session summary data, while the POP sensor only downloads post session.
The Apple Watch is more modern and both the gyro and accel are 3-axis. A free app Sensor Logger allows Apple Watch data to be downloaded and analyzed using signal analysis software.
The signal analysis software analysis shows that it isn't a simple task to analyze tennis shots. Peaks might be positive or negative and of very different amplitudes. However, there does seem to be enough repeatability that the sensors are, overall, reasonably accurate.
Really, that's it?
I actually think so. But, let's try to add another bit of detail. I think it's best to consider the current state of tennis sensors...
Nada.
In the late 2010s, there were at least these tennis sensors:
Sony Smart
Babolat Pop
Babolat Play (integrated tennis racquet)
Zepp Universal
Zepp2 Tennis
Kickstarter that connected to strings*
Another one that I won't link to because it might be for sale but the reviews say it's unsupported
*There is a lot of interesting information about tennis sensors in general if you search for the user SFeingold here.
Why aren't there any more sensors?
You:
No tennis sensors currently exist on the market that I'm aware of. There used to be many. What is your idea of why not?
Copilot:
Tennis sensors have indeed evolved over time, but their presence in the market has fluctuated. Let’s explore some reasons why tennis sensors might not be as prevalent as they once were:
Market Trends and Demand:
- The popularity of tennis sensors can be influenced by market trends and consumer demand. If there’s a surge in interest, manufacturers may invest more in developing and promoting these devices.
- Conversely, if demand wanes or other technologies take center stage, tennis sensors might take a back seat.
Challenges in Accuracy and Consistency:
- Early tennis sensors faced challenges related to accuracy and consistency. Users reported issues with tracking accuracy, especially during fast-paced movements.
- Manufacturers may have struggled to address these technical limitations effectively.
Cost and Affordability:
- Developing high-quality sensors involves research, design, and production costs. If the price point is too high for consumers, adoption rates may suffer.
- Manufacturers need to strike a balance between functionality and affordability.
Competition from Other Wearables:
- Tennis sensors compete with other wearables, such as smartwatches and fitness trackers. Consumers may prioritize multifunctional devices over specialized tennis sensors.
- Brands like Sony, Head, and Babolat have explored sensor technology, but broader wearables offer a wider range of features.
Integration with Smart Court Systems:
- Some tennis venues have adopted Smart Court systems equipped with permanently fixed cameras. These systems record player movements and provide detailed analytics.
- As these systems become more widespread, standalone tennis sensors may face competition.
Innovation and R&D:
- Tennis technology continues to evolve. Manufacturers may be investing in research and development to create more advanced sensors.
- We can hope for better technology down the road as innovation progresses.
Remember that while tennis sensors may not be as prevalent right now, the landscape can change rapidly. Perhaps we’ll see exciting developments in the future! 🎾🔍123
What if people want this information?
The few people that want this information are likely getting it from portable or fixed camera systems. I think one thing that Copilot only touched on, but is perhaps most important, is that it's very difficult to apply tennis data. This website has given ample proof of that. Also, the apps are excellent, but all are missing different features.
The distinction between real-time shot practice and match tracking is something that is only somewhat taken care of with the apps, and long term tracking was hard to do from what I've read. This appears to have made it difficult for most users.
Tennis sensors are dead!
Long live tennis sensors?!?!
I think CoPilot is right to be hopeful, especially with the Apple Watch. Many people have these, so it's just a matter of getting the right app. There already is at least one available. Much of this could be replaced by a camera, but there people might not want to set up a camera each match. There also may be some opportunity to combine visual and sensor data.
Tennis isn't for everybody. Please take lessons if you are interested in swinging the racquet! But, if you decide to, consider tracking your shots with a vision sensor app or an Apple Watch app. Maybe only practice sessions, or maybe only matches. Keeping them separate could increase the information you can get from the sensor.
Don't expect quick results, as many of the best tennis shots are hit softly. But, even though a tennis sensor can't hit shots for you, tracking your shots can be fun and motivating!!
Code?
We need your consent to load the translations
We use a third-party service to translate the website content that may collect data about your activity. Please review the details in the privacy policy and accept the service to view the translations.