Feature Pyramid Networks and Long Short-Term Memory for EEG Feature Map-Based Emotion Recognition

The original EEG data collected are the 1D sequence, which ignores spatial topology information; Feature Pyramid Networks (FPN) is better at small dimension target detection and insufficient feature extraction in the scale transformation than CNN. We propose a method of FPN and Long Short-Term Memory (FPN-LSTM) for EEG feature map-based emotion recognition. According to the spatial arrangement of brain electrodes, the Azimuth Equidistant Projection (AEP) is employed to generate the 2D EEG map, which preserves the spatial topology information; then, the average power, variance power, and standard deviation power of three frequency bands (α, β, and γ) are extracted as the feature data for the EEG feature map. BiCubic interpolation is employed to interpolate the blank pixel among the electrodes; the three frequency bands EEG feature maps are used as the G, R, and B channels to generate EEG feature maps. Then, we put forward the idea of distributing the weight proportion for channels, assign large weight to strong emotion correlation channels (AF3, F3, F7, FC5, and T7), and assign small weight to the others; the proposed FPN-LSTM is used on EEG feature maps for emotion recognition. The experiment results show that the proposed method can achieve Value and Arousal recognition rates of 90.05% and 90.84%, respectively.

High-Density Electromyography Provides Improved Understanding of Muscle Function for Those With Amputation

Transtibial amputation can significantly impact an individual's quality of life including the completion of activities of daily living. Those with lower limb amputations can harness the electrical activity from their amputated limb muscles for myoelectric control of a powered prosthesis. While these devices use residual muscles from transtibial-amputated limb as an input to the controller, there is little research characterizing the changes in surface electromyography (sEMG) signal generated by the upper leg muscles. Traditional surface EMG is limited in the number of electrode sites while high-density surface EMG (HDsEMG) uses multiple electrode sites to gather more information from the muscle. This technique is promising for not only the development of myoelectric-controlled prostheses but also advancing our knowledge of muscle behavior with clinical populations, including post-amputation. The HDsEMG signal can be used to develop spatial activation maps and features of these maps can be used to gain valuable insight into muscle behavior. Spatial features of HDsEMG can provide information regarding muscle activation, muscle fiber heterogeneity, and changes in muscle distribution and can be used to estimate properties of both the amputated limb and intact limb. While there are a few studies that have examined HDsEMG in amputated lower limbs they have been limited to movements such as gait. The purpose of this study was to examine the quadriceps muscle during a slow, moderate and fast isokinetic knee extensions from a control group as well as a clinical patient with a transtibial amputation. HDsEMG was collected from the quadriceps of the dominant leg of 14 young, healthy males (mean age = 25.5 ± 7 years old). Signals were collected from both the intact and amputated limb muscle of a 23 year old clinical participant to examine differences between the affected and unaffected leg. It was found that there were differences between the intact and amputated limb limb of the clinical participant with respect to muscle activation and muscle heterogeneity. While this study was limited to one clinical participant, it is important to note the differences in muscle behavior between the intact and amputated limb limb. Understanding these differences will help to improve training protocols for those with amputation.

Features extraction and multi-classification of sEMG using a GPU-Accelerated GA/MLP hybrid algorithm

BACKGROUND

Surface electromyography (sEMG) signal is the combined effect of superficial muscle EMG and neural electrical activity. In recent years, researchers did large amount of human-machine system studies by using the physiological signals as control signals.

OBJECTIVE

To develop and test a new multi-classification method to improve performance of analyzing sEMG signals based on public sEMG dataset.

METHODS

First, ten features were selected as candidate features. Second, a genetic algorithm (GA) was applied to select representative features from the initial ten candidates. Third, a multi-layer perceptron (MLP) classifier was trained by the selected optimal features. Last, the trained classifier was used to predict the classes of sEMG signals. A special graphics processing unit (GPU) was used to speed up the learning process.

RESULTS

Experimental results show that the classification accuracy of the new method reached higher than 90%. Comparing to other previously reported results, using the new method yielded higher performance.

CONCLUSIONS

The proposed features selection method is effective and the classification result is accurate. In addition, our method could have practical application value in medical prosthetics and the potential to improve robustness of myoelectric pattern recognition.

Functional Information: Towards Synthesis of Biosemiotics and Cybernetics

Biosemiotics and cybernetics are closely related, yet they are separated by the boundary between life and non-life: biosemiotics is focused on living organisms, whereas cybernetics is applied mostly to non-living artificial devices. However, both classes of systems are agents that perform functions necessary for reaching their goals. I propose to shift the focus of biosemiotics from living organisms to agents in general, which all belong to a pragmasphere or functional universe. Agents should be considered in the context of their hierarchy and origin because their semiosis can be inherited or induced by higher-level agents. To preserve and disseminate their functions, agents use functional information - a set of signs that encode and control their functions. It includes stable memory signs, transient messengers, and natural signs. The origin and evolution of functional information is discussed in terms of transitions between vegetative, animal, and social levels of semiosis, defined by Kull. Vegetative semiosis differs substantially from higher levels of semiosis, because signs are recognized and interpreted via direct code-based matching and are not associated with ideal representations of objects. Thus, I consider a separate classification of signs at the vegetative level that includes proto-icons, proto-indexes, and proto-symbols. Animal and social semiosis are based on classification, and modeling of objects, which represent the knowledge of agents about their body (Innenwelt) and environment (Umwelt).

Scaling analysis of paces of fetal breathing, gross-body and extremity movements

Using detrended fluctuation analysis (DFA), we studied the scaling properties of the time instances (occurrence) of the fetal breathing, gross-body, and extremity movements scored on a second by second basis from the recorded ultrasound measurements of 49 fetuses. The DFA exponent α of all the three movements of the fetuses varied between 0.63 and 1.1. We found an increase in α obtained for the movement due to breathing as a function of the gestational age while this trend was not observed for gross-body and extremity movements. This trend was argued as the indication of the maturation of lung and functional development of respiratory aspect of the fetal central nervous system. This result may be useful in discriminating normal fetuses from high-risk fetuses.