Classification of sleep stages is an efficient way of diagnosing sleep problems based on processing the bio-signals (ECG, EEG, EOG, and PPG). The less complex this signal is, the better the detection and processing will be. Feature extraction methods that are done manually are tedious and time-consuming. On the contrary, those features with no hand intervention are called deep features that are usually extracted from images. Analysis of the time-frequency characteristics of non-static bio-signals is of importance since it can provide useful information. The current study aimed to extract the time-frequency image using ECG signal spectrogram as well as the deep features using the convolutional neural network. After extracting the deep features, sleep stages were classified based on deep transfer learning method. Network training was then performed using one of the ECG signals, and testing was done considering the other ECG signal channel. According to the findings, it is possible to detect sleep stages with acceptable accuracy and different amplitudes of signals. Finally, the accuracy and sensitivity values of the sleep stages were measured as 98. 92% and 96. 52%, respectively.

Recently permutation multimedia ciphers were broken in a chosen-plaintext scenario. That attack models a very resourceful adversary which may not always be the case. To show insecurity of these ciphers, we present a cipher-text only attack on speech permutation ciphers. We show inherent redundancies of speech can pave the path for a successful cipher-text only attack. To that end, regularities of speech are extracted in time and frequency using short time Fourier transform. We show that spectrograms of cipher-texts are in fact scrambled puzzles. Then, different techniques including estimation, image processing, and graph theory are fused together in order to create and solve these puzzles. Conducted tests show that the proposed method achieves accuracy of 87: 8% and intelligibility of 92: 9%. These scores are 50: 9% and 34: 6%, respectively, higher than scores of previous method. Finally a novel method, based on moving spectrogram distance, is proposed that can give accurate estimation of segment length of the scrambler system.

Objective: To obtain abnormalities in quantitative Electroencephalography (QEEG) and to observe connectivity between electrodes in children with Asperger disorder.Method: In this study, spectrogram criteria and coherence values are used as a tool for evaluating QEEG in 15 children with Asperger disorder (10 boys and 5 girls aged between 6 to 11 years old) and in 11 control children (7 boys and 4 girls with the same age range).Results: The evaluation of QEEG using statistical analysis and spectrogram criteria demonstrates that the relaxed eye-opened condition in gamma frequency band (34-44Hz) has the best distinction level of 96.2% using spectrogram. The children with Asperger disorder had significant lower spectrogram criteria values (p<0.01) at Fp1 electrode and lower values (p<0.05) at Fp2 and T6 electrodes. Coherence values at 171 pairs of EEG electrodes indicate that the connectivity at (T4, P4), (T4, Cz), (T4, C4) electrode pairs and (T4, O1) had significant differences (p<0.01) in the two groups in the gamma band.Conclusions: It is shown that gamma frequency band can discriminate 96.2% of the two groups using the spectrogram criteria. The results demonstrate that there are more abnormalities in the prefrontal and right temporal lobes using spectrogram criteria and there are more abnormalities in the connectivity of right temporal lobe with the other lobes in the gamma frequency band.

The abstract should include the One of the most exciting topics for researchers over the past few years is detecting underwater acoustic noises. Meanwhile, the complicated nature of the ocean makes this task very challenging. Also, making signals formatted data compatible with machine learning approaches needs much knowledge in signal processing for feature detection. This paper proposed a method to overcome these challenges, which extracts features with Convolutional Neural Network (CNN) and Mel-spectrogram (converting signal data to images). This method needless knowledge in signal processing and more knowledge in machine learning; because using CNNs find the hidden pattern and knowledge of the data automatically. The proposed approach detected the presence of the ships and categorized them into different kinds of them with 99% accuracy that is a noticeable improvement considering state of the art. The performed CNN models consist of 2 CNN layers for feature extraction and a Dense layer for classification the underwater ship noises.

Objective: to evaluate the brain signals in children with autism disorder in many different conditions of quantitative Electroencephalography (qEEG) recordings in order to highlight abnormalities and to characterize this group.Method: In this study, spectrogram was used as a tool for evaluating qEEG in 15 children with autism disorders (13 boys and 2 girls aged between 6 to 11 years old) and in 11 normal children (7 boys and 4 girts with the same age range). signals of the two groups were recorded in nine conditions. Results: The recorded signals with the relaxed eye-opened condition in alpha band, those recorded with looking at a stranger's picture condition in beta band, and the ones obtained with children looking at inverted stranger's picture in the same beta band show the best discrimination of 92.3%, 88,9% and 88.9%respectively using spectrogram. Conclusion: Among the several different EEG recordings, the relaxed eye-opened condition in alpha band had been the best condition for discriminating the two groups using spectrogram. More abnormalities were observed in the prefrontal lobe and the left brain hemisphere in children with autism disorders.

Automatic Speaker Verification (ASV) systems have proven to bevulnerable to various types of presentation attacks, among whichLogical Access attacks are manufactured using voiceconversion and text-to-speech methods. In recent years, there has beenloads of work concentrating on synthetic speech detection, and with the arrival of deep learning-based methods and their success in various computer science fields, they have been a prevailing tool for this very task too. Most of the deep neural network-based techniques forsynthetic speech detection have employed the acoustic features basedon Short-Term Fourier Transform (STFT), which are extracted from theraw audio signal. However, lately, it has been discovered that the usageof Constant Q Transform's (CQT) spectrogram can be a beneficialasset both for performance improvement and processing power andtime reduction of a deep learning-based synthetic speech detection. In this work, we compare the usage of the CQT spectrogram and some most utilized STFT-based acoustic features. As lateral objectives, we consider improving the model's performance as much as we can using methods such as self-attention and one-class learning. Also, short-duration synthetic speech detection has been one of the lateral goals too. Finally, we see that the CQT spectrogram-based model not only outperforms the STFT-based acoustic feature extraction methods but also reduces the processing time and resources for detecting genuine speech from fake. Also, the CQT spectrogram-based model places wellamong the best works done on the LA subset of the ASVspoof 2019 dataset, especially in terms of Equal Error Rate.