Neonatal cry signal prediction and classification via dense convolution neural network

The study of neonatal cry signals is always an interesting topic and still researcher works interminably to develop some module to predict the actual reason for the baby cry. It is really hard to predict the reason for their cry. The main focus of this paper is to develop a Dense Convolution Neural network (DCNN) to predict the cry. The target cry signal is categorized into five class based on their sound as “Eair”, “Eh”, “Neh”, “Heh” and “Owh”. Prediction of these signals helps in the detection of infant cry reason. The audio and speech features (AS Features) were exacted using Mel-Bark frequency cepstral coefficient from the spectrogram cry signal and fed into DCNN network. The systematic DCNN architecture is modelled with modified activation layer to classify the cry signal. The cry signal is collected in different growth phase of the infants and tested in proposed DCNN architecture. The performance of the system is calculated through parameters accuracy, specificity and sensitivity are calculated. The output of proposed system yielded a balanced accuracy of 92.31%. The highest accuracy level 95.31%, highest specificity level 94.58% and highest sensitivity level 93% attain through proposed technique. From this study, it is concluded that the proposed technique is more efficient in detecting cry signal compared to the existing techniques.

Parkinson sEMG signal prediction and generation with Neural Networks

Parkinson’s Disease (PD) is a neurodegenerative disorder characterized by symptoms like resting and action tremors, which cause severe impairments to the patient’s life. Recently, many assistance techniques have been proposed to minimize the disease’s impact on patients’ life. However, most of these methods depend on data from PD’s surface electromyography (sEMG), which is scarce. In this work, we propose the first methods, based on Neural Networks, for predicting, generating, and transferring the style of patient-specific PD sEMG tremor signals. This dissertation contributes to the area by i) comparing different NN models for predicting PD sEMG signals to anticipate resting tremor patterns ii) proposing the first approach based on Deep Convolutional Generative Adversarial Networks (DCGANs) to generate PD’s sEMG tremor signals; iii) applying Style Transfer (ST) for augmenting PD’s sEMG signals with publicly available datasets of non-PD subjects; iv) proposing metrics for evaluating the PD’s signal characterization in sEMG signals. These new data created by our methods could validate treatment approaches on different movement scenarios, contributing to the development of new techniques for tremor suppression in patients.

A Decision-Making Method for Machinery Abnormalities Based on Neural Network Prediction and Bayesian Hypothesis Testing

For anomaly identification of predicted data in machinery condition monitoring, traditional threshold methods have problems during residual testing. It is difficult to make decisions when the residuals are close to the threshold and fluctuate. This paper proposes a Bayesian dynamic thresholding method that combines Bayesian inference with neural network signal prediction. The method makes full use of historical prior data to build an anomaly identification and warning model applicable under single variable or multidimensional variables. A long short-term memory signal prediction model is established, and then a Bayesian hypothesis testing-based anomaly identification strategy is presented to quantify the probability of anomaly occurrence and issue early warnings for anomalies beyond a certain probability. The model was applied to open data sets of a pumping station and actual operating data of a nuclear power turbine. The results indicate that the model successfully predicts the failure probability and failure time. The effectiveness of the proposed method is verified.

Parkinson sEMG signal prediction and generation with Neural Networks

A Doença de Parkinson (DP) é uma doença neurodegenerativa caracterizada por sintomas como tremores de repouso e de ação, que causam graves prejuízos à vida do paciente. Recentemente, diversas dispositivos assistivos têm sido propostos para minimizar o impacto da doença na vida dos pacientes. No entanto, a maioria desses depende de dados da eletromiografia de superfície (sEMG) do paciente, que são escassos. Neste trabalho, propomos os primeiros métodos, baseados em Redes Neurais, para prever e gerar sinais de sEMG de pacientes com Parkinson (PP). Ainda, aprendemos o estilo do tremor no sinal de sEMG destes pacientes, transferindo o mesmo para criar novas amostras a partir de dados de pacientes saudáveis. Esta dissertação contribui para a área i) comparando diferentes modelos para predizer sinais sEMG de pacientes com Parkinson para antecipar padrões de tremor em repouso; ii) propondo a primeira abordagem baseada em Redes Adversariais Generativas Convolucionais Profundas (DCGANs) para gerar sinais de sEMG da doença de Parkinson; iii) aplicando a transferência de estilo (ST) para aumentar o conjunto de sinais sEMG de PPs a partir de dados disponíveis publicamente de indivíduos nãoPP; iv) propondo uma métrica para avaliar a caracterização do sinal de sEMG da doença de Parkinson. Os novos dados criados por nossos métodos podem validar abordagens de tratamento em diferentes cenários de movimento, contribuindo para o desenvolvimento de novas técnicas de supressão de tremor em pacientes.