7.7.13. Human Activity Recognition

Classify human activities from accelerometer and gyroscope data.

7.7.13.1. Overview

This example demonstrates Human Activity Recognition (HAR) using inertial sensor data. It classifies activities such as walking, running, sitting, and standing based on accelerometer and gyroscope readings.

Application: Wearables, fitness trackers, smart home, elderly care

Task Type: Time Series Classification

Data Type: Multivariate (accelerometer + gyroscope)

Note

This example uses a branched model architecture for improved accuracy.

7.7.13.2. Configuration

common:
  target_module: 'timeseries'
  task_type: 'generic_timeseries_classification'
  target_device: 'F28P55'

dataset:
  dataset_name: 'branched_model_parameters'

training:
  model_name: 'CLS_4k_NPU'
  training_epochs: 50
  batch_size: 32

testing: {}
compilation: {}

7.7.13.3. Running the Example

cd tinyml-modelzoo
./run_tinyml_modelzoo.sh examples/branched_model_parameters/config.yaml

7.7.13.4. Dataset Details

Input Variables:

  • Accelerometer X, Y, Z

  • Gyroscope X, Y, Z

Activity Classes:

  • Walking

  • Running

  • Sitting

  • Standing

  • Lying down

  • Walking upstairs

  • Walking downstairs

7.7.13.5. See Also