Machine learning meets Singular Optics II: Single-pixel Detection of Structured Light
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908549793185792 |
|---|---|
| author | Badavath, Purnesh Singh Kumar, Vijay |
| author_facet | Badavath, Purnesh Singh Kumar, Vijay |
| contents | Structured light beams, including Laguerre-Gaussian (LG), Hermite-Gaussian (HG), and perfect vortex (PV) spatial modes, have been at the forefront of modern optics due to their potential in communications, metrology, and sensing. Traditional recognition methods often demand complex alignments and high-resolution imaging. Speckle-learned recognition (SLR) has emerged as a powerful alternative, exploiting the spatio-temporal speckle fields generated by light-diffuser interactions. This paper builds upon the earlier report: Machine Learning Meets Singular Optics (Proc. SPIE 12938, 2024), which demonstrated structured light recognition using 2D speckle images in both on-axis and off-axis channels captured in the spatial domain. In the present work, the recognition framework is advanced by employing 1D speckle information captured in the spatial and temporal domains. This paper reviews how the 2D spatial information of the structured light is mapped on 1D speckle arrays captured in space and 1D temporal speckle fluctuations recorded in time. The 1D speckle arrays captured in the spatial domain have successfully recognised the parent structured light beams with accuracy exceeding 94%, even by employing 1/nth of the 2D speckle data. More recently, 2D spatial information of structured light beams has been mapped onto temporal speckle sequences recorded by a single-pixel detector in the temporal domain. This study highlights the accuracy exceeding 96% across various structured light families, with resilience to turbulence and modal degeneracy. These advances establish scalable, alignment-free, and low-latency recognition architectures suitable for optical communication, sensing, and quantum technologies. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_16946 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Machine learning meets Singular Optics II: Single-pixel Detection of Structured Light Badavath, Purnesh Singh Kumar, Vijay Optics Structured light beams, including Laguerre-Gaussian (LG), Hermite-Gaussian (HG), and perfect vortex (PV) spatial modes, have been at the forefront of modern optics due to their potential in communications, metrology, and sensing. Traditional recognition methods often demand complex alignments and high-resolution imaging. Speckle-learned recognition (SLR) has emerged as a powerful alternative, exploiting the spatio-temporal speckle fields generated by light-diffuser interactions. This paper builds upon the earlier report: Machine Learning Meets Singular Optics (Proc. SPIE 12938, 2024), which demonstrated structured light recognition using 2D speckle images in both on-axis and off-axis channels captured in the spatial domain. In the present work, the recognition framework is advanced by employing 1D speckle information captured in the spatial and temporal domains. This paper reviews how the 2D spatial information of the structured light is mapped on 1D speckle arrays captured in space and 1D temporal speckle fluctuations recorded in time. The 1D speckle arrays captured in the spatial domain have successfully recognised the parent structured light beams with accuracy exceeding 94%, even by employing 1/nth of the 2D speckle data. More recently, 2D spatial information of structured light beams has been mapped onto temporal speckle sequences recorded by a single-pixel detector in the temporal domain. This study highlights the accuracy exceeding 96% across various structured light families, with resilience to turbulence and modal degeneracy. These advances establish scalable, alignment-free, and low-latency recognition architectures suitable for optical communication, sensing, and quantum technologies. |
| title | Machine learning meets Singular Optics II: Single-pixel Detection of Structured Light |
| topic | Optics |
| url | https://arxiv.org/abs/2509.16946 |