Wyniki wyszukiwania - BazTech

1

Using SVM Classifier and Micro-Doppler Signature for Automatic Recognition of Sonar Targets

Saffari Abbas, Zahiri Seye Hamid, Khozein Ghanad Navid

Archives of Acoustics

|

2023

|

Vol. 48, No. 1

49--61

EN

In this paper, we propose using a propeller modulation on the transmitted signal (called sonar micro-Doppler) and different support vector machine (SVM) kernels for automatic recognition of moving sonar targets. In general, the main challenge for researchers and craftsmen working in the field of sonar target recognition is the lack of access to a valid and comprehensive database. Therefore, using a comprehensive mathematical model to simulate the signal received from the target can respond to this challenge. The mathematical model used in this paper simulates the return signal of moving sonar targets well. The resulting signals have unique properties and are known as frequency signatures. However, to reduce the complexity of the model, the 128-point fast Fourier transform (FFT) is used. The selected SVM classification is the most popular machine learning algorithm with three main kernel functions: RBF kernel, linear kernel, and polynomial kernel tested. The accuracy of correctly recognizing targets for different signal-to-noise ratios (SNR) and different viewing angles was assessed. Accuracy detection of targets for different SNRs (−20, −15, −10, −5, 0, 5, 10, 15, 20) and different viewing angles (10, 20, 30, 40, 50, 60, 70, 80) is evaluated. For a more fair comparison, multilayer perceptron neural network with two back-propagation (MLP-BP) training methods and gray wolf optimization (MLP-GWO) algorithm were used. But unfortunately, considering the number of classes, its performance was not satisfactory. The results showed that the RBF kernel is more capable for high SNRs (SNR = 20, viewing angle = 10) with an accuracy of 98.528%.

2

Detection of fillers in the speech by people who stutter

Suszyński Waldemar, Charytanowicz Małgorzata, Rosa Wojciech, Koczan Leopold, Stęgierski Rafał

Applied Computer Science

|

2021

|

Vol. 17, no 4

45--54

EN

Stuttering is a speech impediment that is a very complex disorder. It is difficult to diagnose and treat, and is of unknown initiation, despite the large number of studies in this field. Stuttering can take many forms and varies from person to person, and it can change under the influence of external factors. Diagnosing and treating speech disorders such as stuttering requires from a speech therapist, not only good professional prepa-ration, but also experience gained through research and practice in the field. The use of acoustic methods in combination with elements of artificial intelligence makes it possible to objectively assess the disorder, as well as to control the effects of treatment. The main aim of the study was to present an algorithm for automatic recognition of fillers disfluency in the statements of people who stutter. This is done on the basis of their parameterized features in the amplitude-frequency space. The work provides as well, exemplary results demonstrating their possibility and effectiveness. In order to verify and optimize the procedures, the statements of seven stutterers with duration of 2 to 4 minutes were selected. Over 70% efficiency and predictability of automatic detection of these disfluencies was achieved. The use of an automatic method in conjunction with therapy for a stuttering person can give us the opportunity to objectively assess the disorder, as well as to evaluate the progress of therapy.

3

Feature vector or time-series – comparison of gestures representations in automatic gesture recognition systems

Barczewska K., Wójtowicz W., Moszkowski T.

Challenges of Modern Technology

|

2015

|

Vol. 6, no. 1

1--5

EN

In this paper, we performed recognition of isolated sign language gestures - obtained from Australian Sign Language Database (AUSLAN) – using statistics to reduce dimensionality and neural networks to recognize patterns. We designated a set of 70 signal features to represent each gesture as a feature vector instead of a time series, used principal component analysis (PCA) and independent component analysis (ICA) to reduce dimensionality and indicate the features most relevant for gesture detection. To classify the vectors a feedforward neural network was used. The resulting accuracy of detection ranged between 61 to 87%.

4

Method of automatic recognition and other solutions used in new computer program for full decomposition of EMG signals

Młoźniak D., Piotrkiewicz M.

Biocybernetics and Biomedical Engineering

|

2015

|

Vol. 35, no. 1

22--29

EN

The analysis of electromyographic signals can be very time consuming. In designing a program for EMG signal analysis, there are two competing factors: the accuracy of the final result and its speed. In scientific work, accuracy is the most important factor. All of the existing decomposition programs used in neurophysiology require a final phase of manual corrections, if reliable results are to be obtained. This phase is considerably longer than the phase of automatic recognition. The solutions presented below, used in our new MUR program, allow for the accurate decomposition of complex EMG signals in a reasonable amount of time. The decomposition is performed interactively with optimal time division between automatic and manual tasks. All of this is achieved through a simple method of automatic recognition with the use of the modified coefficient of determination and the method of multiple subtractions of potentials.