Performance of data-driven inner speech decoding with same-task EEG-fMRI data fusion and bimodal models
Authors:
Holly Wilson,
Scott Wellington,
Foteini Simistira Liwicki,
Vibha Gupta,
Rajkumar Saini,
Kanjar De,
Nosheen Abid,
Sumit Rakesh,
Johan Eriksson,
Oliver Watts,
Xi Chen,
Mohammad Golbabaee,
Michael J. Proulx,
Marcus Liwicki,
Eamonn O'Neill,
Benjamin Metcalfe
Abstract:
Decoding inner speech from the brain signal via hybridisation of fMRI and EEG data is explored to investigate the performance benefits over unimodal models. Two different bimodal fusion approaches are examined: concatenation of probability vectors output from unimodal fMRI and EEG machine learning models, and data fusion with feature engineering. Same task inner speech data are recorded from four…
▽ More
Decoding inner speech from the brain signal via hybridisation of fMRI and EEG data is explored to investigate the performance benefits over unimodal models. Two different bimodal fusion approaches are examined: concatenation of probability vectors output from unimodal fMRI and EEG machine learning models, and data fusion with feature engineering. Same task inner speech data are recorded from four participants, and different processing strategies are compared and contrasted to previously-employed hybridisation methods. Data across participants are discovered to encode different underlying structures, which results in varying decoding performances between subject-dependent fusion models. Decoding performance is demonstrated as improved when pursuing bimodal fMRI-EEG fusion strategies, if the data show underlying structure.
△ Less
Submitted 19 June, 2023;
originally announced June 2023.
Multilevel Image Encryption
Authors:
S. Rakesh,
Ajitkumar A Kaller,
B. C. Shadakshari,
B. Annappa
Abstract:
With the fast evolution of digital data exchange and increased usage of multi media images, it is essential to protect the confidential image data from unauthorized access. In natural images the values and position of the neighbouring pixels are strongly correlated. The method proposed in this paper, breaks this correlation increasing entropy of the position and entropy of pixel values using block…
▽ More
With the fast evolution of digital data exchange and increased usage of multi media images, it is essential to protect the confidential image data from unauthorized access. In natural images the values and position of the neighbouring pixels are strongly correlated. The method proposed in this paper, breaks this correlation increasing entropy of the position and entropy of pixel values using block shuffling and encryption by chaotic sequence respectively. The plain-image is initially row wise shuffled and first level of encryption is performed using addition modulo operation. The image is divided into blocks and then block based shuffling is performed using Arnold Cat transformation, further the blocks are uniformly scrambled across the image. Finally the shuffled image undergoes second level of encryption by bitwise XOR operation, and then the image as a whole is shuffled column wise to produce the ciphered image for transmission. The experimental results show that the proposed algorithm can successfully encrypt or decrypt the image with the secret keys, and the analysis of the algorithm also demonstrates that the encrypted image has good information entropy and low correlation coefficients.
△ Less
Submitted 22 February, 2012;
originally announced February 2012.
