A wearable sEMG system must balance signal quality, power use, data volume and synchronization—particularly when multiple muscles are monitored over extended periods. A system developed through the MITOR collaboration between Politecnico di Torino, MIT and Spaulding Rehabilitation Hospital addresses these challenges with six wireless acquisition devices. Each can record conventional surface electromyography (sEMG), a processed signal envelope or an Average Threshold Crossing (ATC) signal, with data stored locally on a microSD card.
The design is relevant to engineers working with distributed sensing and embedded acquisition: it combines an analog front end, digital processing, wireless coordination and local storage, while offering different representations of the same muscle activity. Its evaluation also shows why reduced data volume should be considered alongside the information retained by a signal-processing method.
Wearable sEMG system architecture
Surface electromyography records electrical activity from muscles using electrodes placed on the skin. Unlike intramuscular EMG, it is non-invasive and can be used to observe muscle activation during movement. The recorded signal is an aggregate of activity from recruited motor units and is affected by electrode placement, tissue between muscle and electrode, and noise sources such as motion artifacts, environmental interference and activity from nearby muscles.
For reliable acquisition, an sEMG channel needs to detect the biopotential, amplify and condition it, then digitize it. The MITOR devices add processing options beyond this conventional path. Firmware written in C++ supports three operating modes:
- Classic sEMG: records the sampled signal.
- sEMG envelope: applies digital filtering in firmware to produce an envelope representation.
- ATC: records threshold-crossing activity and calculates an average over a fixed time window.
The design was implemented on a two-layer printed circuit board measuring 50 mm by 37.5 mm. The revised board included an upgraded microcontroller and additional acquisition-channel components for detecting threshold-crossing events. A case was also created to facilitate placement of the devices on subjects during recordings.
How Average Threshold Crossing reduces data
In conventional acquisition, the system samples the analog signal at a defined rate. Threshold Crossing (TC) instead generates an event when the signal exceeds a selected threshold. The resulting output is quasi-digital: the timing between events carries information about the activity, even though the original signal’s waveform is not preserved.
ATC summarizes the number of threshold-crossing events within a fixed time window. This provides a measure related to muscle activity while reducing the volume of data that must be processed or stored compared with recording the complete sEMG waveform. In the described design, threshold detection is supported by acquisition-channel hardware, while firmware manages the operating mode and signal processing.
That reduction comes with a trade-off. A threshold set too low can produce many events; a threshold set too high can omit useful activity. Moreover, TC and ATC do not retain the morphological detail of the original sEMG signal. Engineers therefore need to select the representation according to the analysis objective: waveform detail, an envelope of activity or a compact event-based measure.
Synchronization and operating duration
When multiple devices record different muscles, their data must align in time for meaningful comparison. The MITOR boards use wireless synchronization, enabled by a PCB antenna. Testing reported a maximum synchronization delay of 10 microseconds with a synchronization period of five seconds. This timing capability is important to the system’s intended use in studying coordinated muscle activity across channels.
Power consumption also differed by recording mode. With a 450 mAh battery supply, each developed board was reported to store data for more than 42 hours in classic sEMG mode and more than 80 hours in ATC mode. These figures describe the tested system and its operating modes; they should not be treated as universal battery-life expectations for other wearable devices.
Local microSD storage supports recording without requiring continuous transmission of the full signal. Combined with event-based processing, this approach reduces the data burden associated with acquisition and storage. The design’s central engineering lesson is to evaluate the complete data path—from sensing and processing to storage—rather than treating sampling as an isolated choice.
Validation and muscle synergy analysis
The recorded sEMG was validated against a commercially available multi-channel EMG system from Motion Lab Systems. Tests included upper-limb muscles during grasping and lower-limb muscles during gait. The system was also used to collect data for muscle synergy analysis during walking in a healthy subject.
Muscle synergies describe patterns in which groups of muscles are activated together. In the conventional analysis, processed muscle envelopes are decomposed into muscle-weight patterns and their time-varying activations. The study also tested whether ATC-derived data could support synergy extraction. The comparison assessed zero-lag cross-correlation between time activations and cosine similarity between muscle weights; the reported ATC results showed good agreement with the conventional approach in this evaluation.
This comparison does not mean that ATC preserves all the information available in raw sEMG. Rather, it indicates that a lower-data representation can be useful for a particular analysis task when its outputs are evaluated against the conventional method.
Practical implications for embedded monitoring
For engineers designing wearable or distributed acquisition platforms, the system highlights several decisions that should be addressed together:
- Choose the signal representation deliberately. Raw sEMG, an envelope and ATC provide different balances between detail and data volume.
- Design synchronization into the architecture. Multi-channel analysis depends on aligning measurements from separate devices.
- Evaluate thresholds as part of system performance. Threshold selection affects event frequency and the information represented.
- Measure power in the intended modes. Operating mode can materially affect the reported recording duration.
- Validate against an established measurement system. Comparisons help assess signal quality and downstream analysis suitability.
Conclusion
The MITOR wearable sEMG system brings together multi-device synchronization, local recording and three signal modes for muscle-activity monitoring. Its ATC approach reduces data volume by representing activity as threshold events and windowed counts, but sacrifices waveform detail and depends on suitable threshold selection. For technical teams, the practical takeaway is to match the acquisition method to the analysis goal, then validate timing, signal quality, data handling and power use as parts of one integrated system.