A Brain Computer Interface Based on FFT and multilayer neural network - feature extraction and generalization -

Nakayama, Kenji; Kaneda, Yasuaki; Hirano, Atsushi

doi:10.1109/ispacs.2007.4446015

Cited by 21 publications

(17 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, the generalization methods have been applied to the MLNN based BCI. The method of adding small random numbers to the MLNN input data can improve classification performance [13].…”

mentioning

confidence: 99%

See 1 more Smart Citation

A BCI using MEGvision and multilayer neural network - channel optimization and main lobe contribution analysis -

Nakayama

Kaneda

Hirano

et al. 2009

2008 International Symposium on Intelligent Signal Processing and Communications Systems

View full text Add to dashboard Cite

Abstract-Multilayer neural networks (MLNN) and the FFT amplitude of brain waves have been applied to 'Brain ComputerInterface' (BCI). In this paper, a magnetoencephalograph (MEG) system, 'MEGvision' developed by Yokogawa Corporation, is used to measure brain activities. MEGvision is a 160-channel whole-head MEG system. Channels are selected from 8 main regions, a frontal lobe, a temporal lobe, a parietal lobe and a occipital lobe, located in the left and the right sides of the brain. The 8 channels, located at the central point in the 8 lobes, are initially selected. Optimum channels are searched for in the same lobe as the initial channels in order to achieve high classification accuracy. Two subjects and four mental tasks, including relaxed situation, multiplication, playing sport and rotating an object, are used. The brain waves are measured 10 times for one subject and one mental task. Among them, 8 data sets are used for training the MLNN, and the remaining 2 data sets are used for testing. 5 kinds of combinations of 2 data sets are selected for testing. Rates of correct classification by using the initial channels are 82.5 ∼ 90%. By optimizing the channels, the accuracy is improved up to 85.0 ∼ 97.5%, which is very high accuracy. Furthermore, contributions of the brain waves in the 8 lobes are analyzed.

show abstract

mentioning

confidence: 99%

“…In this paper, the BCI system based on the FFT amplitude and the MLNN is employed [12], [13]. The magnetoencephalograph (MEG) system, 'MEGvision' developed by Yokogawa Corporation, is used to measure brain activities.…”

mentioning

confidence: 99%

A BCI using MEGvision and multilayer neural network - channel optimization and main lobe contribution analysis -

Nakayama

Kaneda

Hirano

et al. 2009

2008 International Symposium on Intelligent Signal Processing and Communications Systems

View full text Add to dashboard Cite

show abstract

“…For this reason, the nonlinear normalization as shown in Eq. (1) has been introduced [17]. x is the FFT amplitude before normalization and f (x) is the normalized amplitude.…”

Section: Nonlinear Normalizationmentioning

confidence: 99%

“…In our previous work, two kinds of generalization techniques, which are adding small random numbers to the MLNN input data [14] and a weight decay technique [15], have been applied. The former method can provide good classification performance [17].…”

Section: Generalization By Adding Small Random Numbersmentioning

confidence: 99%

See 1 more Smart Citation

Neural Network Based BCI by Using Orthogonal Components of Multi-channel Brain Waves and Generalization

Nakayama

Horita

Hirano

Artificial Neural Networks - ICANN 2008

View full text Add to dashboard Cite

Abstract. FFT and Multilayer neural networks (MLNN) have been applied to 'Brain Computer Interface' (BCI). In this paper, in order to extract features of mental tasks, individual feature of brain waves of each channel is emphasized. Since the brain wave in some interval can be regarded as a vector, Gram-Schmidt orthogonalization is applied for this purpose. There exists degree of freedom in the channel order to be orthogonalized. Effect of the channel order on classification accuracy is investigated. Next, two channel orders are used for generating the MLNN input data. Two kinds of methods using a single NN and double NNs are examined. Furthermore, a generalization method, adding small random numbers to the MLNN input data, is applied. Simulations are carried out by using the brain waves, available from the Colorado State University website. By using the orthogonal components, a correct classification rate Pc can be improved from 70% to 78%, an incorrect classification rate Pe can be suppressed from 10% to 8%. As a result, a rate Rc = Pc/(Pc + Pe) can be improved from 0.875 to 0.907. When two different channel orders are used, Pe can be drastically suppressed from 10% to 2%, and Rc can be improved up to 0.973. The generalization method is useful especially for using a sigle channel order. Pc can be increased up to 84 ∼ 88% and Pe can be suppressed down to 2 ∼ 4%, resulting in Rc = 0.957 ∼ 0.977.

show abstract

A Review on Algorithms for EEG-Based BCIs

Das

Tripathy

Raheja

2018

SpringerBriefs in Applied Sciences and Technology

View full text Add to dashboard Cite

Many classes of electrophysiological activities of the brain are used in designing various types of brain-computer interfaces (BCIs). Those are discussed briefly in this section. Sensorimotor activity generally corresponds to the behavior of the brain rhythms (mu, beta, and gamma), movement-related potentials (MRPs), etc. Next, the classification of BCI based on various parameters has also been discussed. These parameters are the mode of signal acquisition, timing, and placement of sensors. Later in this section, algorithms that have been and with chances of being, used in BCI applications have been discussed in a detailed manner. The algorithms chosen for each stage of the signal processing have an equal role to play in resulting a better outcome. Therefore, the section emphasizes the algorithms for each such stage, separately. Choosing a perfect algorithm is very important to design an efficient classifier. This section provides important information about the algorithms concerned with BCI. Sensorimotor ActivitySensorimotor cortex is the source of mu rhythms (ranging from 8 to 12 Hz) and beta rhythms (ranging from 13 to 30 Hz). These are generated when a person is not involved in sensory or motor activities. They are most distinct in the frontal and parietal lobes of the brain and make changes in the mu and lower beta bands. Any voluntary movement in the subject increases the power in the brain rhythms or frequencies. This phenomenon is also known as event-related synchronization (ERS). The peak of an ERS occurs at a delay of 600 ms following the movement offset. Finally, gamma rhythm is a high-frequency rhythm in the electroencephalography (EEG). The amplitude of gamma rhythm increases upon the occurrence of a movement.Movement-related potentials (MRPs) usually have a maximum amplitude at the vertex with low-frequency potentials, which become more significant close to the movement. Sensorimotor activities other than the previously mentioned ones are not restricted to any particular band of frequencies or location in the brain. For example,

show abstract

A Brain Computer Interface Based on FFT and multilayer neural network - feature extraction and generalization -

Cited by 21 publications

References 12 publications

A BCI using MEGvision and multilayer neural network - channel optimization and main lobe contribution analysis -

A BCI using MEGvision and multilayer neural network - channel optimization and main lobe contribution analysis -

Neural Network Based BCI by Using Orthogonal Components of Multi-channel Brain Waves and Generalization

A Review on Algorithms for EEG-Based BCIs

Contact Info

Product

Resources

About