An Ensemble One Dimensional Convolutional Neural Network with Bayesian Optimization for Environmental Sound Classification

Speaker-based language identification for Ethio-Semitic languages using CRNN and hybrid features

Natural language is frequently employed for information exchange between humans and computers in modern digital environments. Natural Language Processing (NLP) is a basic requirement for technological advancement in the field of speech recognition. For additional NLP activities like speech-to-text translation, speech-to-speech translation, speaker recognition, and speech information retrieval, language identification (LID) is a prerequisite. In this paper, we developed a Language Identification (LID) model for Ethio-Semitic languages. We used a hybrid approach (a convolutional recurrent neural network (CRNN)), in addition to a mixed (Mel frequency cepstral coefficient (MFCC) and mel-spectrogram) approach, to build our LID model. The study focused on four Ethio-Semitic languages: Amharic, Ge'ez, Guragigna, and Tigrinya. By using data augmentation for the selected languages, we were able to expand our original dataset of 8 h of audio data to 24 h and 40 min. The proposed selected features, when evaluated, achieved an average performance accuracy of 98.1%, 98.6%, and 99.9% for testing, validation, and training, respectively. The results show that the CRNN model with (Mel-Spectrogram + MFCC) combination feature achieved the best results when compared to other existing models.

Towards Automated Optimization of Residual Convolutional Neural Networks for Electrocardiogram Classification

The interpretation of biological data such as the ElectroCardioGram (ECG) signal gives clinical information and helps to assess the heart function. There are distinct ECG patterns associated with a specific class of arrhythmia. The convolutional neural network, inspired by findings in the study of biological vision, is currently one of the most commonly employed deep neural network algorithms for ECG processing. However, deep neural network models require many hyperparameters to tune. Selecting the optimal or the best hyperparameter for the convolutional neural network algorithm is a highly challenging task. Often, we end up tuning the model manually with different possible ranges of values until a best fit model is obtained. Automatic hyperparameters tuning using Bayesian Optimization (BO) and evolutionary algorithms can provide an effective solution to current labour-intensive manual configuration approaches. In this paper, we propose to optimize the Residual one Dimensional Convolutional Neural Network model (R-1D-CNN) at two levels. At the first level, a residual convolutional layer and one-dimensional convolutional neural layers are trained to learn patient-specific ECG features over which multilayer perceptron layers can learn to produce the final class vectors of each input. This level is manual and aims to limit the search space and select the most important hyperparameters to optimize. The second level is automatic and based on our proposed BO-based algorithm. Our optimized proposed architecture (BO-R-1D-CNN) is evaluated on two publicly available ECG datasets. Comparative experimental results demonstrate that our BO-based algorithm achieves an optimal rate of 99.95% for the MIT-BIH database to discriminate between five kinds of heartbeats, including normal heartbeats, left bundle branch block, atrial premature, right bundle branch block, and premature ventricular contraction. Moreover, experiments demonstrate that the proposed architecture fine-tuned with BO achieves a higher accuracy tested on the 10,000 ECG patients dataset compared to the other proposed architectures. Our optimized architecture achieves excellent results compared to previous works on the two benchmark datasets.

Number of Convolution Layers and Convolution Kernel Determination and Validation for Multilayer Convolutional Neural Network: Case Study in Breast Lesion Screening of Mammographic Images

Mammography is a low-dose X-ray imaging technique that can detect breast tumors, cysts, and calcifications, which can aid in detecting potential breast cancer in the early stage and reduce the mortality rate. This study employed a multilayer convolutional neural network (MCNN) to screen breast lesions with mammographic images. Within the region of interest, a specific bounding box is used to extract feature maps before automatic image segmentation and feature classification are conducted. These include three classes, namely, normal, benign tumor, and malignant tumor. Multiconvolution processes with kernel convolution operations have noise removal and sharpening effects that are better than other image processing methods, which can strengthen the features of the desired object and contour and increase the classifier’s classification accuracy. However, excessive convolution layers and kernel convolution operations will increase the computational complexity, computational time, and training time for training the classifier. Thus, this study aimed to determine a suitable number of convolution layers and kernels to achieve a classifier with high learning performance and classification accuracy, with a case study in the breast lesion screening of mammographic images. The Mammographic Image Analysis Society Digital Mammogram Database (United Kingdom National Breast Screening Program) was used for experimental tests to determine the number of convolution layers and kernels. The optimal classifier’s performance is evaluated using accuracy (%), precision (%), recall (%), and F1 score to test and validate the most suitable MCNN model architecture.