DenseNet Based Speech Imagery EEG Signal Classification using Gramian Angular Field

Islam, Md. Monirul; Shuvo, Md. Maruf Hossain

doi:10.1109/icaee48663.2019.8975572

Cited by 20 publications

(12 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the following, the dimensions (see: figure 1) of our benchmark comparison are explained in further detail. We highlight that the signal acquisition and data pre-processing of the first step (i.e., data foundation) vary greatly depending on the underlying domain and classification problem (e.g., dimensionality reduction [31], wavelet transform [35]). Therefore, targeting a cross-domain investigation of visualized raw data from different datasets, we focus on identifying relevant domains and adequate datasets in the first dimension [36].…”

Section: Methodsmentioning

confidence: 99%

“…By assessing related work concerning these three dimensions, it can be seen that studies are either restricted to one dataset, belonging to one isolated domain [35], [38], only base their findings on a single visualization technique [39], or are built upon a small sample size [31], manifesting limited validity and do not allow to expand proposed approaches to classification problems with an insufficient sample size [34]. As studies have shown that using non-raw data and including manual feature extraction is highly time-consuming [17] and includes the risk of wrongfully classifying relevant features in the time-series data [12], [16], the state-of-research manifests high potential for improvement of objectivity and efficiency.…”

Section: Related Workmentioning

confidence: 99%

“…TL greatly reduces the amount of required training data by utilizing knowledge acquired for a related domain to solve a task for another, improving the model's performance and robustness [33], [34]. Although the current time-series imaging literature lacks TL-based approaches for optimized model performance, frameworks from different domains (e.g., healthcare [35], engineering [31]) allow subclassifying the CNN-based classification of visualized time-series data into three main steps across domains [30], [31], [35], [36]: 1) Data foundation (i.e., signal acquisition and data preprocessing) 2) Imaging (i.e., encoding time-series as images) 3) Evaluation (i.e., DL model for label classification) Companies can use these three influential dimensions (see: figure 1) to gain competitive advantages over their competitors by optimizing their approaches for data analysis [29]. By using benchmarking as an established tool for comparison [37], companies can identify best practices for performance improvement.…”

Section: Introductionmentioning

confidence: 99%

“…Studies are either adapted to a specific domain [38], only use a single visualization technique [39], manifest a small sample size [31], or manifest the theoretical issue of long-term dependencies as their approaches are based on RNNs [40], [41]. Therefore, related work is limited either theoretically [40], [41] or practically [35], [38], [39] regarding the systematic analysis of raw time-series data across domains.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Benchmarking Transfer Learning Strategies in Time-Series Imaging: Recommendations for Analyzing Raw Sensor Data

2022

View full text Add to dashboard Cite

With the growing availability and complexity of time-series sequences, scalable and robust machine learning approaches are required that overcome the sampling challenge of quantitatively sufficient training data. Following the research trend towards the deep learning-based analysis of time-series encoded as images, this study proposes a time-series imaging workflow that overcomes the challenge of quantitatively limited sensor data across domains (i.e., medicine and engineering). After systematically identifying the three relevant dimensions that affect the performance of the deep learning-based analysis of visualized timeseries data, we performed a benchmarking evaluation with a total of 24 unique convolutional neural network models. Following a two-level transfer learning investigation, we reveal that fine-tuning the mid-level features results in the best classification performance. As a result, we present an optimized representation of the VGG16 network, which outperforms previous studies in the field. Our approach is accurate, robust, and manifests internal and external validity. By only using the raw time-series data, our model does not require manual feature engineering, being of high practical relevance. As the post-hoc analysis of our results reveals that our model allows automated extraction of meaningful features based on the trend of the underlying timeseries data, our study also adds to explainable artificial intelligence. Furthermore, our proposed workflow reduces the sequence length of the input data while preserving all information. Especially with the hurdle of long-term dependencies in sequential time-series data, we overcome related work's limitation of the vanishing gradients problem and contribute to the sequential learning theory in artificial intelligence.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Benchmarking Transfer Learning Strategies in Time-Series Imaging: Recommendations for Analyzing Raw Sensor Data

2022

View full text Add to dashboard Cite

show abstract

“…These researchers are part of an active research community investigating the potential of non-invasive brain measurement technologies for speech neuroprostheses. Techniques include scalp-electroencephalography (EEG) [45][46][47][48][49][50][51] , providing high temporal resolution, especially the Kara One database 52 provides the foundation for many studies; Magnetoencephalography 53,54 , providing more localized information than EEG; and Functional Near Infrared Spectroscopy [55][56][57][58] , providing localized information of cortical hemoglobin levels. The advances made by this community could also benefit invasive speech neuroprostheses, a dataset that is provided to everyone could be used to evaluate and leverage their approaches.…”

mentioning

confidence: 99%

Speech Production in Intracranial Electroencephalography: iBIDS Dataset

Verwoert

Ottenhoff

Colon

et al. 2022

Preprint

View full text Add to dashboard Cite

Speech production is an intricate process involving a large number of muscles and cognitive processes. The neural processes underlying speech production are not completely understood. As speech is a uniquely human ability, it can not be investigated in animal models. High-fidelity human data can only be obtained in clinical settings and is therefore not easily available to all researchers. Here, we provide a dataset of 10 participants reading out individual words while we measured intracranial EEG from a total of 1103 electrodes. The data, with its high temporal resolution and coverage of a large variety of cortical and sub-cortical brain regions, can help in understanding the speech production process better. Simultaneously, the data can be used to test speech decoding and synthesis approaches from neural data to develop speech Brain-Computer Interfaces and speech neuroprostheses.

show abstract