Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimizing TinyML: The Impact of Reduced Data Acquisition Rates for Time Series Classification on Microcontrollers

Riya Samanta, Bidyut Saha, Soumya K. Ghosh, Ram Babu Roy

TL;DR

Reducing data acquisition rates affects TinyML models for time series classification, focusing on resource-constrained, battery operated IoT devices, and offers valuable insights for deploying efficient TinyML models in constrained environments.

Abstract

Tiny Machine Learning (TinyML) enables efficient, lowcost, and privacy preserving machine learning inference directly on microcontroller units (MCUs) connected to sensors. Optimizing models for these constrained environments is crucial. This paper investigates how reducing data acquisition rates affects TinyML models for time series classification, focusing on resource-constrained, battery operated IoT devices. By lowering data sampling frequency, we aim to reduce computational demands RAM usage, energy consumption, latency, and MAC operations by approximately fourfold while maintaining similar classification accuracies. Our experiments with six benchmark datasets (UCIHAR, WISDM, PAMAP2, MHEALTH, MITBIH, and PTB) showed that reducing data acquisition rates significantly cut energy consumption and computational load, with minimal accuracy loss. For example, a 75\% reduction in acquisition rate for MITBIH and PTB datasets led to a 60\% decrease in RAM usage, 75\% reduction in MAC operations, 74\% decrease in latency, and 70\% reduction in energy consumption, without accuracy loss. These results offer valuable insights for deploying efficient TinyML models in constrained environments.

Optimizing TinyML: The Impact of Reduced Data Acquisition Rates for Time Series Classification on Microcontrollers

TL;DR

Reducing data acquisition rates affects TinyML models for time series classification, focusing on resource-constrained, battery operated IoT devices, and offers valuable insights for deploying efficient TinyML models in constrained environments.

Abstract

Tiny Machine Learning (TinyML) enables efficient, lowcost, and privacy preserving machine learning inference directly on microcontroller units (MCUs) connected to sensors. Optimizing models for these constrained environments is crucial. This paper investigates how reducing data acquisition rates affects TinyML models for time series classification, focusing on resource-constrained, battery operated IoT devices. By lowering data sampling frequency, we aim to reduce computational demands RAM usage, energy consumption, latency, and MAC operations by approximately fourfold while maintaining similar classification accuracies. Our experiments with six benchmark datasets (UCIHAR, WISDM, PAMAP2, MHEALTH, MITBIH, and PTB) showed that reducing data acquisition rates significantly cut energy consumption and computational load, with minimal accuracy loss. For example, a 75\% reduction in acquisition rate for MITBIH and PTB datasets led to a 60\% decrease in RAM usage, 75\% reduction in MAC operations, 74\% decrease in latency, and 70\% reduction in energy consumption, without accuracy loss. These results offer valuable insights for deploying efficient TinyML models in constrained environments.
Paper Structure (19 sections, 2 figures, 8 tables)

This paper contains 19 sections, 2 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Background
  3. TinyML and Resource Constraints
  4. Time Series Classification
  5. Data Acquisition Rate
  6. Methodology
  7. Datasets
  8. UCIHAR
  9. WISDM
  10. PAMAP2
  11. MHEALTH
  12. MIT-BIH
  13. PTB Diagnostic ECG Database
  14. Data Preprocessing
  15. Model Architecture
  16. ...and 4 more sections

Figures (2)

  • Figure 1: Time series signals from various sensors (e.g., accelerometer, gyroscope, magnetometer, temperature, humidity, presence sensors etc) are processed by TinyML models running directly on MCUs (such as ESP32, Raspberry Pi Pico, BLE 33 Nano, NRF52840 etc).
  • Figure 2: Radar plots comparing various metrics (Accuracy, FLASH (KB), RAM (KB), MACs (K), Time (ms), Energy (µJ)) across different data acquisition rate reduction levels: 0% (blue), 25% (green), 50% (red), and 75% (purple). Each plot is normalized for better comparison and shows performance and resource utilization trade-offs for different datasets. The shaded areas illustrate how accuracy and efficiency vary with reduction levels. This visualization highlights that significant reductions in data acquisition rate can lead to a lower resource requirement while maintaining similar levels of accuracy.