Vibration measurements are essential to verify mechanical stability, quantify resonance behaviour, early failure detection, etc. Existing approaches include high-speed camera–based imaging that capture motion by tracking visual features (e.g., edges/markers), but it requires expensive high-frame-rate cameras and high-resolution optics. Other approaches include contact sensors, which provide highly quantitative, low-noise measurements, yet they are inherently single-point and require physical contact, being more suitable for permanent installations. Comparable high-precision data can be obtained with the use of a Laser Doppler Vibrometer. Although it is essentially a point-measurement method, a two-dimensional displacement map can be generated by scanning the sample surface with the laser beam. This process is time consuming and incapable of capturing simultaneous, full-field transient phenomena.
A method that circumvents these limitations is laser vibrometry based on the analysis of speckle patterns. In this approach, a larger volume of, e.g., a sample under study is illuminated by an expanded coherent laser beam. Illuminating the rough sample surface generates a random interference phenomenon known as a speckle pattern. This pattern is recorded with a camera. When the target surface vibrates, this leads to dynamically changing speckle pattern fluctuations. By analysing these fluctuations, it is possible to reconstruct the underlying vibration characteristics, specifically frequency and amplitude.
Historically, these analyses have mainly been carried out using high-speed CMOS cameras. However, these systems are inherently limited by integration times and frame rate constraints, which often prove insufficient for high-frequency micro-vibrations. While techniques such as laser speckle contrast imaging reduce these limitations in the reconstruction of the vibration amplitude, they fail to provide a robust solution for frequency recovery due to temporal averaging. Typically, images are taken with exposure times longer than the vibration period which leads to a blur of the speckle pattern. The amplitude then can be estimated via the local speckle contrast reduction.
Recently, event-based cameras, also known as neuromorphic cameras, have emerged as a disruptive alternative. Unlike frame-based sensors, event cameras detect asynchronous, per-pixel brightness changes, offering microsecond-level temporal resolution and a high dynamic range. This architecture eliminates redundant static data and motion blur, capturing the precise timing of pixel intensity changes events. In combination with laser speckle illumination, event‑based cameras enable high‑temporal‑resolution vibration measurements. With these cameras, it is possible to reconstruct vibration frequencies up to tens of kilohertz at very low amplitudes.
In this work, we compare the performance of laser-speckle vibration detection via conventional CMOS camera and event-based camera across a range of controlled vibration frequencies and amplitudes. We evaluate the performance of the different algorithms in the reconstruction of vibration frequency and amplitude, from frame-based image analysis to event-based asynchronous event stream analysis. Our results demonstrate the limitations of both technologies and illustrate potential of event-based technology overcoming the vibration frequency bandwidth and frame-rate constraints of conventional CMOS architectures, offering a more efficient and precise solution for vibration monitoring with significantly reduced data overhead.