Into the latent space of capacitive sensors: interpolation and synthetic data generation using variational autoencoders

M. Monteagudo Honrubia, F.J. Herraiz-Martínez, J. Matanza

Machine Learning: Science and Technology Vol. 6, nº. 1, pp. 015031-1 - 015031-15

Summary:

For many sensing applications, collecting a large experimental dataset could be a time-consuming and expensive task that can also hinder the implementation of Machine Learning models for analyzing sensor data. Therefore, this paper proposes the generation of synthetic signals through a Variational Autoencoder (VAE) to enlarge a spectra dataset acquired with a capacitive sensor based on a Dielectric Resonator. Trained with signals of several water/glycerine concentrations, this generative model learns the dataset characteristics and builds a representative latent space. Consequently, exploring this latent space is a critical task to control the generation of synthetic signals and interpolating concentrations unmeasured by the sensor. For this reason, this paper proposes a search method based on Bayesian Optimization that automatically explores the latent space. The results show excellent signal reconstruction quality, proving that the VAE architecture can successfully generate realistic synthetic signals from capacitive sensors. In addition, the proposed search method obtains a reasonable interpolation capability by finding latent encodings that generate signals related to the target glycerin concentrations. Moreover, this approach could be extended to other sensing technologies.

Spanish layman's summary:

Recopilar grandes conjuntos de datos adquiridos por sensores puede resultar costoso en tiempo y recursos, lo que limita el desarrollo de modelos de aprendizaje automático en aplicaciones de detección. Nuestra última investigación presenta una solución innovadora: el uso de Variational Autoencoders (VAE) para generar señales sintéticas realistas para sensores capacitivos. Estos modelos aprenden las características clave del conjunto de datos y puede crear nuevas señales a partir de ellas, incluso para concentraciones no medidas. Además, se desarrolló un método de optimización bayesiana para explorar el espacio latente y guiar el proceso de generación.

English layman's summary:

Collecting large datasets for sensor-based applications can be expensive and time-consuming, limiting the development of Machine Learning models. Our latest research introduces an innovative solution: using Variational Autoencoders (VAEs) to generate realistic synthetic signals for capacitive sensors. These models learn the key characteristics of the dataset and can create new, high-quality data points—even for unmeasured concentrations. In addition, we developed a Bayesian search method that explores the VAE latent space to guide the generation process.

Keywords: variational autoencoders, latent space search, data augmentation, synthetic data, generative models, capacitive sensors, class interpolation

JCR Impact Factor and WoS quartile: 6,300 - Q1 (2023)

DOI reference: https://doi.org/10.1088/2632-2153/adb009

Published on paper: March 2025.

Published on-line: February 2025.

Citation:
M. Monteagudo Honrubia, F.J. Herraiz-Martínez, J. Matanza, Into the latent space of capacitive sensors: interpolation and synthetic data generation using variational autoencoders. Machine Learning: Science and Technology. Vol. 6, nº. 1, pp. 015031-1 - 015031-15, March 2025. [Online: February 2025]