KEYWORDS: Data storage, Image segmentation, Holography, Deep learning, Holographic data storage systems, Neural networks, Education and training, Data modeling, Mathematical optimization, Holograms
Experiments have shown that deep learning can improve the data reading of holographic data storage. However, it requires a large amount of storage materials and time to obtain data to optimize the network model. In data encoding, each encoded data page consists of 51sub-pages with the same structure. This paper proposes a deep learning method for image segmentation based on encoding features in collinear holographic data storage. Using a deep learning method of image segmentation, the encoded data page is segmented into data sub-pages. It can reduce material loss and data collection time.
Although polarization holography introduces polarization dimensions, it is well known that polarization has only two orthogonal dimensions, and the expansion of recording capabilities is limited. Therefore, we introduce the polarization encoding for theoretical analysis and calculation, the orthogonal polarization array of arbitrary dimensions is obtained. Assuming that the n-dimensional vectors Q1, Q2, …, and Qx are a group of non-zero vectors that are orthogonal to each other in the orthogonal polarization array. The Schmidt orthogonalization method is used to expand the column vector group of the n-dimensional orthogonal polarization array into a set of canonical orthogonal basis of the space Kn. During the experiment, when the signal S1 is recorded with Q1, it can be faithfully reconstructed with Q1, while it shows null reconstruction with Q2 or Qx. By analogy, multiple recording and independent reconstruction experiments are carried out successively.
Now the era of big data has arrived, and there is an urgent need for high storage capacity storage solutions to store large amounts of data. As a new generation of storage technology, holographic optical storage has the advantages of large data storage capacity, fast transmission speed, read-write parallelism and so on. The storage material for holographic storage should have the characteristics of fast response, high signal-to-noise ratio, high diffraction efficiency and high stability. Phenanthrenequinone-doped poly (methyl methacrylate) (PQ/PMMA) photopolymer is a common storage material, which has the advantages of high diffraction efficiency, inexpensive and simple preparation. Currently, PQ/PMMA is mainly prepared manually. The reproducibility of the preparation process faces challenge due to human errors. Therefore, we designed an automatic PQ/PMMA preparation device, which can effectively eliminate the differences caused by human factors. We have verified through experiments that materials prepared automatically have better stability than those prepared manually. Among the prepared single sheet materials by automatic preparation device, we measured that the difference in diffraction efficiency at different positions is within 10%. The automated experimental platform provides assistance for the stable preparation of materials
With the rapid development of information technology, the amount of data has shown explosive growth. The traditional magnetic storage and optical storage can no longer gradually meet the needs of data storage. Holographic data storage breaks through the mode of two-dimensional data storage and stores data in the form of three-dimensional volume, which can improve the data storage density by one dimension and bring ultra-fast data transfer rate at the same time. However, to promise holographic data storage work well, the servo system should be used in practice to avoid the effect of vibration.
Holographic data storage is a powerful potential technology to solve the problem of mass data long-term storage. To increase the storage capacity, the information to be stored is encoded into a complex amplitude. Fast and accurate retrieval of amplitude and phase from the reconstructed beam is necessary during data readout. In this talk, we propose a complex amplitude demodulation method based on deep learning from a single-shot diffraction intensity image and verified it by a non-interferometric lensless experiment demodulating four-level amplitude and four-level phase. By analyzing the correlation between the diffraction intensity features and the amplitude and phase encoding data pages, the inverse problem is decomposed into two backward operators denoted by two convolutional neural networks to demodulate amplitude and phase respectively. The stable and simple complex amplitude demodulation and strong anti-noise performance from the deep learning provide an important guarantee for the practicality of holographic data storage.
KEYWORDS: Data modeling, Signal to noise ratio, Deep learning, Holographic data storage systems, Holography, Data storage, Education and training, Objectives, Neural networks, Reliability
In recent years, optical holographic data storage system has gradually become a research hotspot and a strong competitor of big data storage due to its high data transfer rate, long storage life and high storage density. In the collinear amplitude modulated holographic data storage system, in order to improve the storage density, a high magnification objective lens is usually used as the recording lens to record the encoded data pages in the holographic storage medium. Therefore, when the objective lens is focused on the holographic storage medium, the accuracy and reliability of data recording and reading can be guaranteed. However, in the process of normal use of the system, environmental interference and other factors will inevitably lead to defocusing of the objective lens, which will result in high bit-error-rate (BER) and low signal-to-noise ratio (SNR) of the recorded and read coding information, affecting the accuracy and reliability of information reading. In this paper, we propose a collinear amplitude modulated holographic data storage system objective defocusing correction model using deep learning. Only a training model with defocusing distance of 100μm can be used to correct the defocusing of the objective lens with defocusing distance less than 100μm. The reconstructed BER is reduced to less than 1/10 of the original data, and the SNR is increased to more than 5 times of the original data. The reliability and accuracy of system record reading are improved.
Compared with traditional iterative methods, deep learning phase reconstruction has lower bit error rate and higher data transfer rate. We found the efficiency of training mainly was from the edges of the phase patterns due to their stronger intensity changes between adjacent phase distribution. According to this characteristic, we proposed a method to only record and use the high frequency component of the phase patterns and to do the deep learning training. This method can improve the storage density due to reducing the material consumption.
The phase retrieval method based on deep learning can be used to solve the iterative problem in holographic data storage. The key of the deep learning method is to build the relationship between the phase data pages and the corresponding near-field diffraction intensity patterns. However, to build the correct relationship, thousands of samples of the training dataset are usually required. In this paper, according to the coding characteristics of phase data pages, we proposed an image segmentation method to greatly reduce the number of original training dataset. The innovation proposed by this new method lies in the special segmentation of the original samples to expand the number of samples.
Holographic data storage is one powerful potential technology to solve the problem of mass data long-term storage. Deep learning is showing its advantages in many fields such as artificial intelligence, detection and imaging. When deep learning meets holographic data storage, new modulation ways and decoding methods were born. We did three kinds of modulation amplitude only, phase only and complex amplitude respectively in holographic data storage and used deep learning method to do data reconstruction. The results were better than previous reconstruction methods. Data reconstruction based on deep learning owns more anti-noise performance.
Amplitude-modulated collinear holographic data storage technology has high storage density, fast data transfer rate and stable system. The key to realizing system operation is to decode the amplitude code quickly and correctly. We proposed a decoding method based on 3:16 amplitude code. We used the convolution calculation to locate the sync mark point of every sub-page in the data page quickly and calculated the magnification rates among sub-pages to get the correct sub-page image segmentation. Taking the bit error rate as the evaluation standard, we verified our method successfully in different image quality.
Polarization holography has great potential in Ultra-high-definition (UHD) information diplay and data storage. Due to the faithful reconstruction in polarization holography, the storage capacity is further improved easily. In this paper, a device for generating vector vortex beam is demonstrated using the faithful reconstruction characteristics. Through the analysis of the experimental results, it is found that the helical phase order corresponding to different polarization states is different in the transmission process. It shows the independence of vector vortex beam propagation. This method has a certain research space in optical storage, and application prospect in optical micromanipulation optical tweezers.
Phase-modulated holographic data storage shows a great prospect in Ultra-high-definition (UHD) information display and data storage due to its higher capacity than amplitude modulation. However, the phase reconstruction is more sensitive to noise in the spectrum plane. In this paper, we proposed to use the low-depth camera to obtain the spectral intensity of the reconstructed beam, and used iterative Fourier transform algorithm to retrieve phase. Simulation and experiment show that this method has stronger noise suppression performance.
We used the amplitude coding method of 3:16, that is, in a 4 * 4 pixel matrix, only three pixels are in the on state, and the remaining pixels are in the off state. In the collinear amplitude holographic data storage system, U-Net full convolution neural network is used to denoise the amplitude coded image obtained by the detector. The experimental results show that the bit error rate can be reduced to less than 1% from 10% and the image signal-to-noise ratio can be increased by more than 5 times.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.