A Survey on FloraVoice: TinyML and IoT Based Acoustic Stress Monitoring in Plants

Abstract

Agricultural monitoring systems are transitioning from reactive, environment-centric approaches toward proactive, plant-centric frameworks. FloraVoice exploits a less commonly studied signal source: acoustic activity produced by plants under physiological stress. When water supply is interrupted, especially during drought-induced cavitation, short ultrasonic events may be produced. This survey examines IoT plant monitoring, soil sensing, agricultural ultrasound, TinyML-based audio classifi- cation, and automated irrigation systems to understand how such acoustic evidence can be applied in a practical setting. The paper describes a FloraVoice framework using piezoelectric sensors, noise cancellation, analog filtering, ESP32-based signal processing, and edge inference, and concludes that plant acoustics can strengthen ordinary soil and climate sensing by adding a plant-response layer, although reliable deployment requires care- ful handling of noise, calibration, labelled datasets, processing overhead, and model size.

Introduction

The text introduces FloraVoice, a proposed smart agricultural monitoring system that uses plant-generated ultrasonic acoustic signals, IoT sensors, and TinyML (Tiny Machine Learning) to detect plant stress at an earlier stage than conventional methods.

Background and Motivation

Traditional plant monitoring systems rely on visible symptoms such as wilting, discoloration, or reduced growth to identify stress. However, by the time these signs appear, the plant may have already suffered significant damage. Modern IoT-based agriculture has improved monitoring through sensors that measure soil moisture, temperature, and humidity, but these systems mainly observe environmental conditions rather than the plant's actual physiological response.

Research suggests that stressed plants can emit ultrasonic acoustic signals caused by internal processes such as xylem cavitation during drought. Detecting these signals could provide much earlier warnings of stress before visible symptoms occur.

Role of TinyML

TinyML enables machine learning models to run directly on low-power microcontrollers. In FloraVoice:

• Acoustic signals are processed locally on the device.
• Compact spectral features are extracted.
• A trained model classifies plant stress in real time.
• Only alerts or final results are transmitted, reducing communication and power consumption.

This makes the system inexpensive, energy-efficient, and suitable for large-scale deployment.

Literature Survey Findings

The reviewed studies covered several related areas:

1. IoT Smart Agriculture Systems
   • Monitor soil moisture, temperature, and humidity.
   • Improve irrigation management.
   • Limitation: do not directly measure plant stress.
2. Agricultural Sensor Technologies
   • Include moisture, pH, nutrient, and wearable plant sensors.
   • Lack integrated acoustic stress-detection pipelines.
3. TinyML Acoustic Event Detection
   • Demonstrates low-power edge AI for sound classification.
   • Not specifically designed for plant ultrasound.
4. Agricultural Ultrasound Monitoring
   • Shows ultrasound can reveal plant health conditions.
   • Faces challenges such as noise and signal interpretation.
5. Smart Irrigation Systems
   • Use environmental sensors for automated watering.
   • Depend heavily on connectivity and indirect measurements.
6. Plant Ultrasound Studies
   • Confirm that plants emit detectable ultrasonic signals under stress.
   • Limited by the availability of large, labeled datasets.

Existing System Limitations

Most current low-cost agricultural monitoring systems:

• Measure only environmental factors.
• Use fixed threshold-based decisions.
• Depend on cloud connectivity.
• Cannot directly observe plant physiological responses.
• Often detect stress only after significant damage has occurred.

These systems may also struggle to adapt to different plant species, growth stages, and environmental conditions.

Problem Statement

The primary challenge is the delayed detection of plant stress. Existing monitoring approaches fail to identify internal stress early enough because they rely on environmental measurements or visible symptoms.

An ideal system should:

• Detect stress before visible damage occurs.
• Monitor plant responses directly.
• Be low-cost and energy-efficient.
• Operate locally without constant cloud dependence.
• Be scalable for farms, greenhouses, and home gardening.

Proposed System: FloraVoice

FloraVoice combines:

• Plant ultrasonic sensing
• Environmental monitoring
• TinyML-based edge intelligence

Key components include:

1. Dual Piezoelectric Sensors
   • One sensor attached to the plant stem captures stress-related ultrasonic signals.
   • A second reference sensor captures environmental vibrations.
2. Signal Processing
   • A high-pass filter removes background noise.
   • Signals are digitized using an ESP32-WROOM-32 microcontroller.
3. TinyML Classification
   • Acoustic features are extracted.
   • A machine learning model identifies stress events locally.
4. Real-Time Alerts
   • Early stress warnings are generated without waiting for visible symptoms.

Conclusion

This survey reviewed IoT, sensor, ultrasound, and TinyML research relevant to FloraVoice, a plant acoustic stress moni- toring system. Existing plant monitoring systems work well for environmental observation but consistently detect stress only after external conditions or visible symptoms indicate a problem. Acoustic emissions offer a complementary signal that may reveal internal plant stress at an earlier stage and with greater specificity. The proposed FloraVoice architecture combines piezoelec- tric sensing, analog filtering, ESP32-based digital signal pro- cessing, and TinyML classification to detect possible stress events at the edge. Future work should focus on building a labelled plant acoustic dataset that covers multiple species and noise conditions, validating the model across real field deployments, improving noise robustness through advanced filtering techniques, and integrating acoustic alerts with au- tomated irrigation control for closed-loop water management.

References

[1] S. Daad, R. A. Rahman, and M. H. A. Zainal, “IoT-Based Smart Agriculture Monitoring System,” in Proc. Natl. Conf. Educ.—Tech. & Vocational Training, 2025. [2] W. Song et al., “Soil Sensors in Smart Agriculture: Multi-Type Monitor- ing Technologies and Ecological Development Approaches,” Agriculture, vol. 16, no. 3, 2026, doi: 10.3390/agriculture16030159. [3] B. B. Roy, S. Das, and U. K. Mondal, “TinyML-Driven Sensor Nodes for Energy-Efficient Acoustic Event Detection in Pervasive Acoustic WSNs,” [4] J. Telecommun. Inf. Technol., vol. 100, no. 2, pp. 69–77, 2025, doi: 10.26636/jtit.2025.2.2086. [5] M. A. Sarkar and D. S. Laila, “A Review of Ultrasound Monitoring Applications in Agriculture,” Front. Plant Sci., vol. 16, 2025, doi: 10.3389/fpls.2025.1620868. [6] R. Geetha, R. Rakhi, and K. N. Anith, “Innovative Sensor Technologies in Agriculture—Applications and Challenges: A Review,” Agric. Sci. Digest, 2025, doi: 10.18805/ag.D-6344. [7] S. M. Charave et al., “Smart Plant Monitoring and Automated Irrigation System Using IoT,” Int. J. Adv. Res. Sci. Commun. Technol., vol. 5, no. 4, 2023, doi: 10.48175/IJARSCT-26481. [8] I. Bonnati, I. Pelacha, and L. Arru, “Plant Ultrasound Detection: A Cost- Effective Method for Identifying Plant Ultrasonic Emissions,” Plant Sig- nal. Behav., vol. 19, no. 1, 2024, doi: 10.1080/15592324.2024.2310874. [9] J. Nie et al., “Real-Time Multisource Fusion of Satellite-UAV Remote Sensing Images and Acoustic Signals for Crop Monitoring,” IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 2025.

Copyright

Copyright © 2026 Avilash Rout, Chandre Gowda S N, Prem Kumar R, Punith Raj H, Prof. Hemalatha K. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.