7.8.20. Dynamic Hand Gesture Recognition
Real-time dynamic hand gesture classification from 3-axis accelerometer data on MSPM0G5187 with integrated NPU.
7.8.20.1. Overview
Task: Multi-class classification (4 gesture types)
Application: Gesture-based human-machine interface
Dataset: 3-axis accelerometer recordings from TI Sensor BoosterPack
Device: MSPM0G5187 (NPU-accelerated)
7.8.20.2. Gesture Classes
Circle — Clockwise or counter-clockwise circular motion
Wave — Side-to-side waving motion
Tap — Short impact gesture
Others — Non-gesture or unrecognized motion
7.8.20.3. Running the Example
cd tinyml-modelzoo
./run_tinyml_modelzoo.sh examples/dynamic_hand_gesture_recognition/config_MSPM0.yaml
cd tinyml-modelzoo
run_tinyml_modelzoo.bat examples\dynamic_hand_gesture_recognition\config_MSPM0.yaml
7.8.20.4. Device Support
Device |
Hardware |
Configuration File |
|---|---|---|
|
MSPM0 with NPU + TI Sensor BoosterPack (3-axis accelerometer) |
|
7.8.20.5. Configuration
common:
task_type: 'generic_timeseries_classification'
target_device: 'MSPM0G5187'
dataset:
dataset_name: 'Hand_gesture_dataset'
input_data_path: 'https://software-dl.ti.com/C2000/esd/mcu_ai/01_04_00/datasets/hand_gesture_dataset.zip'
data_processing_feature_extraction:
feature_extraction_name: 'Input256_RAW_256Feature_1Frame_3InputChannel_removeDC_2D1'
variables: 3
training:
model_name: 'CLS_55k_NPU'
training_epochs: 40
batch_size: 30
quantization: 2
quantization_method: 'QAT'
quantization_weight_bitwidth: 8
quantization_activation_bitwidth: 8
testing:
enable: True
compilation:
enable: True
7.8.20.6. Dataset Description
Sensor: 3-axis accelerometer (X, Y, Z) from TI Sensor BoosterPack
Variables: X, Y, Z acceleration (
variables: 3)Frame size: 256 samples per frame
Stride: 0.25 (64-sample step between frames)
Normalization: Range normalization per frame
7.8.20.7. Feature Extraction
The example uses raw time-domain features (no FFT). Range normalization captures gesture amplitude and shape directly.
7.8.20.8. Available Models
Model |
Parameters |
Description |
|---|---|---|
|
~55,000 |
Default — CNN optimized for NPU, best accuracy |
7.8.20.9. Expected Results
Accuracy: ~94.46% on test data
Device: MSPM0G5187 NPU
7.8.20.10. System Components
Hardware
MSPM0G5187 microcontroller with integrated NPU
TI Sensor BoosterPack with 3-axis accelerometer
Software
Code Composer Studio 12.x or later
MSPM0 SDK 2.10.04 or later
7.8.20.11. Next Steps
Learn about feature extraction: Feature Extraction
Deploy to device: NPU Device Deployment
Browse similar examples: Human Fall Detection, Human Activity Recognition