Gaussian Mixture Model Classifier Research Articles

A semiparametric Gaussian mixture model for chest CT-based 3D blood vessel reconstruction.

Computed tomography (CT) has been a powerful diagnostic tool since its emergence in the 1970s. Using CT data, 3D structures of human internal organs and tissues, such as blood vessels, can be reconstructed using professional software. This 3D reconstruction is crucial for surgical operations and can serve as a vivid medical teaching example. However, traditional 3D reconstruction heavily relies on manual operations, which are time-consuming, subjective, and require substantial experience. To address this problem, we develop a novel semiparametric Gaussian mixture model tailored for the 3D reconstruction of blood vessels. This model extends the classical Gaussian mixture model by enabling nonparametric variations in the component-wise parameters of interest according to voxel positions. We develop a kernel-based expectation-maximization algorithm for estimating the model parameters, accompanied by a supporting asymptotic theory. Furthermore, we propose a novel regression method for optimal bandwidth selection. Compared to the conventional cross-validation-based (CV) method, the regression method outperforms the CV method in terms of computational and statistical efficiency. In application, this methodology facilitates the fully automated reconstruction of 3D blood vessel structures with remarkable accuracy.

Impact of florbetaben acquisition timing on continuous and dichotomous measures of amyloid burden: implications for real‐world amyloid PET interpretation

AbstractBackgroundThe magnitude of Standardized Uptake Value Ratios (SUVRs) and Centiloid (CL) values increase at later acquisition windows for Florbetaben (FBB). As amyloid PET is anticipated to be utilized for treatment decisions with anti‐amyloid therapeutics (van Dyck, 2022), it is important to understand acquisition factors that influence PET quantification.MethodWe analyzed 120 FBB scans from the Stanford Alzheimer’s Disease Research Center with 5‐minute frame data spanning 70‐110 minutes post‐injection (Table 1). Frame data were motion corrected, summed into 70‐90 and 90‐110‐minute files, and processed through an MRI‐Free pipeline. A global cortical target region and whole cerebellum reference region from the GAAIN atlas were used to create SUVRs and CLs (Klunk, 2014). SUVRs and CLs computed across the two acquisition windows were compared using linear regression to understand the effects of timing acquisition on magnitude. Gaussian Mixture Models (GMM) were used to convert SUVRs to a probability scale separately for 70‐90 and 90‐110 datasets. 10,000 bootstrapped mixed timing datasets were created by randomly selecting either 70‐90 or 90‐110 SUVRs for a given participant, which was then used to understand the impact of GMM‐based classification into A‐ and A+ groups when acquisition time varies within dataset.ResultCLs were systematically lower during the earlier acquisition window (2.01 CL for the CL<10 group, ΔCL range = ‐8.83,9.16; 2.89 CL for the 10‐25 group, ΔCL range = ‐2.18,9.33; 3.62 CL for the 25‐50 group, ΔCL range = ‐0.04,11.34; and 10.94 for the >50 group, ΔCL range = 1.81,26.98) (Fig. 1). GMM classification into A‐ and A+ groups was consistent across the two acquisition windows and in the mixed timing datasets (Table 2).ConclusionGMM is a robust quantitative method that can be used to characterize amyloid PET scans into A‐ and A+ groups when acquisition time varies. However, given the inconsistent impact of acquisition time on CL magnitude, comparison of CL values before and after anti‐amyloid treatments will require adjustment for scan acquisition time. Future work is needed to establish methods that can measure amyloid PET change when acquisition times vary within an individual patient. This is especially important for evaluating the efficacy of currently approved anti‐amyloid therapeutic approaches.

Impact of florbetaben acquisition timing on continuous and dichotomous measures of amyloid burden: implications for real‐world amyloid PET interpretation

AbstractBackgroundThe magnitude of Standardized Uptake Value Ratios (SUVRs) and Centiloid (CL) values increase at later acquisition windows for Florbetaben (FBB). As amyloid PET is anticipated to be utilized for treatment decisions with anti‐amyloid therapeutics (van Dyck, 2022), it is important to understand acquisition factors that influence PET quantification.MethodWe analyzed 120 FBB scans from the Stanford Alzheimer’s Disease Research Center with 5‐minute frame data spanning 70‐110 minutes post‐injection (Table 1). Frame data were motion corrected, summed into 70‐90 and 90‐110‐minute files, and processed through an MRI‐Free pipeline. A global cortical target region and whole cerebellum reference region from the GAAIN atlas were used to create SUVRs and CLs (Klunk, 2014). SUVRs and CLs computed across the two acquisition windows were compared using linear regression to understand the effects of timing acquisition on magnitude. Gaussian Mixture Models (GMM) were used to convert SUVRs to a probability scale separately for 70‐90 and 90‐110 datasets. 10,000 bootstrapped mixed timing datasets were created by randomly selecting either 70‐90 or 90‐110 SUVRs for a given participant, which was then used to understand the impact of GMM‐based classification into A‐ and A+ groups when acquisition time varies within dataset.ResultCLs were systematically lower during the earlier acquisition window (2.01 CL for the CL<10 group, ΔCL range = ‐8.83,9.16; 2.89 CL for the 10‐25 group, ΔCL range = ‐2.18,9.33; 3.62 CL for the 25‐50 group, ΔCL range = ‐0.04,11.34; and 10.94 for the >50 group, ΔCL range = 1.81,26.98) (Fig. 1). GMM classification into A‐ and A+ groups was consistent across the two acquisition windows and in the mixed timing datasets (Table 2).ConclusionGMM is a robust quantitative method that can be used to characterize amyloid PET scans into A‐ and A+ groups when acquisition time varies. However, given the inconsistent impact of acquisition time on CL magnitude, comparison of CL values before and after anti‐amyloid treatments will require adjustment for scan acquisition time. Future work is needed to establish methods that can measure amyloid PET change when acquisition times vary within an individual patient. This is especially important for evaluating the efficacy of currently approved anti‐amyloid therapeutic approaches.

Malicious attack detection based on continuous Hidden Markov Models in Wireless sensor networks

Alongside the rapid progress of Wireless sensor networks (WSNs) technologies, sensors and networks can rapidly be a victim of distributed attacks. Attackers can perform intrusions to break down the network during the routing process and intercept gathered data by dropping them for example. To prevent the increase of security issues in WSN, many attack identification methods and models were proposed in which detection systems are deployed to collect sensed data and categorize them using machine learning and stochastic binary-classification techniques. For this work, a new method is introduced to analyze and classify an own WSN dataset. We aim to design an anomaly identification approach to improve the sensor network security and its efficiency with high accuracy. To reach this goal, machine learning approaches are used to define a detection system that learns from routing datasets to identify network malicious entries. The proposed models are based on Hidden Markov Model (HMM), and Gaussian Mixture Model (GMM) stochastic assumptions. Also, the dimensionality reduction technique was deployed to select the most relevant features for the training. The experimentation phase was realized on its own and made the dataset that reflects different network situations of normal and attacked cases. The outcomes performances of the proposed method were obtained with a classification accuracy of 92.18% using a 2 HMM/3 GMM classifier. This result demonstrates the quality of our proposed approach compared with existing literature and its usefulness to improve the security of WSN.

Hybrid Strategies for Efficient Intra Prediction in Spatial SHVC

With multi-layer encoding and Inter-layer prediction, Spatial Scalable High Efficiency Video Coding (SSHVC) has extremely high coding complexity. It is very crucial to speed up its coding to promote widespread and cost-effective SSHVC applications. Specifically, we first reveal that the average RD cost of Inter-layer Reference (ILR) mode is different from that of Intra mode, but they both follow the Gaussian distribution. Based on this discovery, we apply the classic Gaussian Mixture Model and Expectation Maximization to determine whether ILR mode is the best mode thus skipping Intra mode. Second, when coding units (CUs) in enhancement layer use Intra mode, it indicates very simple texture is presented. We investigate their Directional Mode (DM) distribution, and divide all DMs into three classes, and then develop different methods with respect to classes to progressively predict the best DMs. Third, by jointly considering rate distortion costs, residual coefficients and neighboring CUs, we propose to employ the Conditional Random Fields model to early terminate depth selection. Experimental results demonstrate that the proposed algorithm can significantly improve coding speed with negligible coding efficiency losses.

Alcoholic EEG Signal Classification Using Multi-Heuristic Classifiers with Stochastic Gradient Descent Technique for Tuning the Hyperparameters

Electroencephalography (EEG) signals are utilized to examine various pathological as well as physiological brain activities. Alcoholism is an example of a significant behavior that may be investigated and comprehended utilizing electrical brain impulse models. In the area of biomedical research, categorizing alcoholic patients using EEG data is a complicated issue. To overcome this issue, in this research, alcoholic EEG signal classification was performed by various dimensionality reduction techniques like Hilbert Transform, Rigid Regression and Chi Square Probability Density Function. Finally, the Bayesian Linear Discriminant Classifier, Linear Regression, Logistic Regression, Gaussian Mixture Model (GMM), Adaboost, Detrend Fluctuation Analysis, Firefly Algorithm, Harmonic Search Algorithm, and Cuckoo Search Algorithm are employed to classify the dimensionally reduced alcoholic EEG dataset. In addition, we provide an approach for selecting the ideal combination of Stochastic Gradient Descent (SGD)-based hyper parameters updation algorithm to improve the accuracy of alcoholic EEG classification in GMM, Firefly, Harmonic Search, and Cuckoo Search classifiers in this study. When dimensionally reduced alcoholic EEG signal features from the Hilbert Transform are used with the SGD with GMM classifier, the results display good accuracy of 96.31%.

EGMM: An evidential version of the Gaussian mixture model for clustering

The Gaussian mixture model (GMM) provides a simple yet principled framework for clustering, with properties suitable for statistical inference. In this paper, we propose a new model-based clustering algorithm, called EGMM (evidential GMM), in the theoretical framework of belief functions to better characterize cluster-membership uncertainty. With a mass function representing the cluster membership of each object, the evidential Gaussian mixture distribution composed of the components over the powerset of the desired clusters is proposed to model the entire dataset. The parameters in EGMM are estimated by a specially designed Expectation–Maximization (EM) algorithm. A validity index allowing automatic determination of the proper number of clusters is also provided. The proposed EGMM is as simple as the classical GMM, but can generate a more informative evidential partition for the considered dataset. The synthetic and real dataset experiments show that the proposed EGMM performs better than other representative clustering algorithms. Besides, its superiority is also demonstrated by an application to multi-modal brain image segmentation.

Epileptic Seizures Detection from EEG Recordings Based on a Hybrid system of Gaussian Mixture Model and Random Forest Classifier

Epilepsy is the most common neurological disease defined as a central nervous system disorder that is characterized by recurrent seizures. While electroencephalography (EEG) is an essential tool for monitoring epilepsy patients' brain activity and diagnosing epilepsy, Visual detection of the EEG signal to identify epileptic seizures is a time-consuming approach that might result in human error. Therefore, an early and precise epilepsy diagnosis is critical to reducing the risk of future seizures. This paper aims to increase epileptic seizure detection accuracy in a balanced dataset while reducing the execution time. To address this, we proposed a hybrid system of supervised and unsupervised machine learning algorithms to construct a computationally efficient and scalable model for the early detection of epileptic seizures from two-class EEG datasets. First, Discrete Wavelet Transform (DWT) was applied to the EEG signal to decompose it into frequency sub-bands. Then these EEG extracted features were fed into the Gaussian Mixture Model (GMM) for partitioning these features into two clusters: epilepsy or not. Lastly, the clusters' output was evaluated with the random forest classifier. In addition, Principal Component Analysis (PCA) was used to reduce the EEG features and to reduce further the features obtained after conducting DWT on the EEG signal to determine the impacts of dimension reduction on this system performance. The experimental results show that the highest accuracy was achieved by the hybrid system of GMM with random forest with DWT features with an accuracy of 93.62 %.

Gaussian Mixture Model classifier analog integrated low-power implementation with applications in fault management detection

An Integrated Analog Gaussian Mixture Model classifier architecture is introduced consisting of multiple Gaussian function circuits and a Winner-Take-All circuit. It is modular and scalable to the number of classes and clusters, and, to the input dimensionality. The operating principles of the classifier are illustrated in detail and are used in a low-power, low-voltage and fully-tunable implementation targeting bearing fault management applications. The implementation was done in a 90nm CMOS process using the Cadence IC Suite for the electrical and physical design. Post-layout simulation results were compared with a software implementation of the classifier confirming the proper operation of the design.

Performance Exploration of Multiple Classifiers with Grid Search Hyperparameter Tuning for Detecting Epileptic Seizures from EEG Signals

This study evaluates the performance of two-level classifications using dimensionality reduction methods to determine the risk level of epilepsy from EEG dataset. To diminish the complexity of EEG data, dimensionality reduction techniques such as Singular Value Decomposition (SVD), Independent Component Analysis (ICA), and Principal Component Analysis (PCA) are utilized. The risk level of epilepsy classification from EEG dataset would then be carried out using three classifiers: Hidden Markov Model (HMM), Naïve Bayesian Classifier (NBC) and Gaussian Mixture Model (GMM). The Grid Search (GS) process is employed to tune the hyperparameters of GMM and NBC classifiers. This study analyzed twenty patients’ datasets. Performance evaluation of classifiers with and without GS hyperparameter tuning is examined, including performance index, sensitivity, specificity, and accuracy. The GMM classifier with the GS hyper-tuning approach for SVD dimensionality reduction technique achieved a higher accuracy of 98.18% than its counterpart classifiers.

Detection of liver abnormalities—A new paradigm in medical image processing and classification techniques

AbstractThe liver is the body's most essential organ, and all human activities are interrelated with normal liver function. Any malfunction of the liver may lead to fatal diseases; therefore, early detection of liver abnormalities is essential. Modern medical imaging techniques combined with engineering procedures are reducing human suffering caused by liver disease. This study uses multiple classifiers to detect liver cirrhosis in ultrasonic images. The ultrasound images were obtained from The Cancer Imaging Archive database. A gray‐level co‐occurrence matrix (GLCM) and statistical approaches are used to extract features from normal and liver‐cirrhosis images. The extracted GLCM features are normalized and classified using nonlinear regression, linear regression, logistic regression, Bayesian Linear Discriminant Classifiers (BLDC), Gaussian Mixture Model (GMM), Firefly, Cuckoo search, Particle Swarm Optimization (PSO), Elephant search, Dragon Fly, Firefly GMM, Cuckoo search GMM, PSO GMM, Elephant search GMM, and Dragon Fly GMM classifiers. Benchmark metrics, such as sensitivity, specificity, accuracy, precision, negative predictive value, false‐negative rate, balanced accuracy, F1 score, Mathew correlation coefficient, F measure, error rate, Jaccard metric, and classifier success index, are assessed to identify the best‐performing classifier. The GMM classifier outperformed other classifiers for statistical features, and it achieved the highest accuracy (98.39%) and lowest error rate (1.61%). Moreover, the Dragon Fly GMM classifier achieved 90.69% for the GLCM feature used to classify liver cirrhosis.

On significance of constant-Q transform for pop noise detection

Liveness detection has emerged as an important research issue for many biometrics, such as face, iris, hand geometry, etc. and significant research efforts are reported in the literature. However, less emphasis is given to liveness detection for voice biometrics or Automatic Speaker Verification (ASV). Voice Liveness Detection (VLD) can be a potential technique to detect spoofing attacks in ASV system. Presence of pop noise in the speech signal of live speaker provides the discriminative acoustic cue to distinguish between genuine vs. spoofed speech in the framework of VLD. Pop noise comes out as a burst at the lips, which is captured by the ASV system (since the speaker and microphone are close enough), indicating the liveness of the speaker and provides the basis of VLD. In this paper, we present the Constant-Q Transform (CQT) -based approach over the traditional Short-Time Fourier Transform (STFT) -based algorithm (baseline). With respect to Heisenberg’s uncertainty principle in signal processing framework, the CQT has variable spectro-temporal resolution, in particular, better frequency resolution for low frequency region and better temporal resolution for high frequency region, which can be effectively utilized to identify the low frequency characteristics of pop noise. We have also compared proposed algorithm with cepstral features, namely, Linear Frequency Cepstral Coefficients (LFCC) and Constant-Q Cepstral Coefficients. The experiments are performed on recently released POp noise COrpus (POCO) dataset with various statistical, discriminative, and deep learning-based classifiers, namely, Gaussian Mixture Model (GMM), Support Vector Machine (SVM), Convolutional Neural Networks (CNN), Light-CNN (LCNN), and Residual Network (ResNet), respectively. The significant improvement in performance, in particular, an absolute improvement of 14.23% and 10.95% in terms of percentage classification accuracy on development and evaluation set, respectively, is obtained for the proposed CQT-based algorithm along with SVM classifier, over the STFT-SVM (baseline) system. Similar trend of the performance improvement is observed for the GMM, CNN, LCNN, and ResNet classifiers for the proposed CQT-based algorithm vs. traditional STFT-based algorithm. The analysis is further extended by simulating the replay mechanism (in the standard framework of ASVSpoof-2019 PA challenge dataset) on the subset of POCO dataset in order to observe the effect of room acoustics onto the performance of the VLD system. By embedding the moderate simulated replay mechanism in POCO dataset, we obtained the percentage classification accuracy of 97.82% on evaluation set.

Classifying Kepler light curves for 12 000 A and F stars using supervised feature-based machine learning

ABSTRACT With the availability of large-scale surveys like Kepler and TESS, there is a pressing need for automated methods to classify light curves according to known classes of variable stars. We introduce a new algorithm for classifying light curves that compares 7000 time-series features to find those that most effectively classify a given set of light curves. We apply our method to Kepler light curves for stars with effective temperatures in the range 6500–10 000 K. We show that the sample can be meaningfully represented in an interpretable 5D feature space that separates seven major classes of light curves (δ Scuti stars, γ Doradus stars, RR Lyrae stars, rotational variables, contact eclipsing binaries, detached eclipsing binaries, and non-variables). We achieve a balanced classification accuracy of 82 per cent on an independent test set of Kepler stars using a Gaussian mixture model classifier. We use our method to classify 12 000 Kepler light curves from Quarter 9 and provide a catalogue of the results. We further outline a confidence heuristic based on probability density to search our catalogue and extract candidate lists of correctly classified variable stars.

Unintentional modulation microstructure enlargement

Radio frequency fingerprinting (RFF) is a technology that identifies the specific emitter of a received electromagnetic signal by external measurement of the minuscule hardware-level, device-specific imperfections. The RFF-related information is mainly in the form of unintentional modulation (UIM), which is subtle enough to be effectively imperceptible and is submerged in the intentional modulation (IM). It is necessary to minimize the influence of the IM and expand the slight differences between emitters for successful RFF. This paper proposes a UIM microstructure enlargement (UMME) method based on feature-level adaptive signal decomposition (ASD), accompanied by autocorrelation and cross-correlation analysis. The common IM part is evaluated by analyzing a newly-defined benchmark feature. Three different indexes are used to quantify the similarity, distance, and dependency of the RFF features from different devices. Experiments are conducted based on the real-world signals transmitted from 20 of the same type of radar in the same working mode. The visual image qualitatively shows the magnification of feature differences; different indicators quantitatively describe the changes in features. Compared with the original RFF feature, recognition results based on the Gaussian mixture model (GMM) classifier further validate the effectiveness of the proposed algorithm.

Comparing Machine Learning Algorithms in Land Use Land Cover Classification of Landsat 8 (OLI) Imagery

In recent times, there have been increased rates at which researchers are searching for advanced ways of carrying out land-use land-cover (LULC) mapping, especially in developing countries. Four machine-learning algorithms, namely Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbour (K-NN), and Gaussian Mixture Models (GMM) were examined. This study also attempted to validate the various models using the index-based validation method. Accuracy assessment was performed by using the Kappa coefficient. The results of the LULC showed that RF classified 23% of the study area as bare land, SVM has 24% of the study area classified as bare land, K-NN also allotted 24% to bare land, while that of GMM classifier was 30%. The overall accuracy of RF, SVM, K-NN and GMM were 0.9840, 0.9780, 0.9641 and 0.9421 respectively. The Kappa Coefficient of the various classifiers were RF (0.9695), SVM (0.9580), K-NN (0.9319) and GMM (0.8916). This study showed that though all the algorithms performed relatively very well, RF performed better than the other classifiers. It suffices to state that, there is a need for further studies since other extraneous environmental variables may be underpinning these conclusions.

Nanopower Integrated Gaussian Mixture Model Classifier for Epileptic Seizure Prediction

This paper presents a new analog front-end classification system that serves as a wake-up engine for digital back-ends, targeting embedded devices for epileptic seizure prediction. Predicting epileptic seizures is of major importance for the patient’s quality of life as they can lead to paralyzation or even prove fatal. Existing solutions rely on power hungry embedded digital inference engines that typically consume several W or even mW. To increase the embedded device’s autonomy, a new approach is presented combining an analog feature extractor with an analog Gaussian mixture model-based binary classifier. The proposed classification system provides an initial, power-efficient prediction with high sensitivity to switch on the digital engine for the accurate evaluation. The classifier’s circuit is chip-area efficient, operating with minimal power consumption (180 nW) at low supply voltage ( V), allowing long-term continuous operation. Based on a real-world dataset, the proposed system achieves 100% sensitivity to guarantee that all seizures are predicted and good specificity (69%), resulting in significant power reduction of the digital engine and therefore the total system. The proposed classifier was designed and simulated in a TSMC 90 nm CMOS process, using the Cadence IC suite.

The impact of sensing parameters on data management and anomaly detection in structural health monitoring

The massive and autonomous structural health monitoring (SHM) of bridges is a problem that is of growing interest due to its importance and topicality. However, a considerable amount of data must be elaborated and managed in such an application. This paper proposes a set of machine learning (ML) tools to detect anomalies in a bridge from vibrational measurements using the minimum amount of data. The proposed framework starts from the fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by a density-based time-domain tracking algorithm. The fundamental frequencies extracted are then fed to one-class classification (OCC) algorithms that perform anomaly detection. Then, to reduce the amount of data, we analyze the effect of the number of sensors, the number of bits per sample, the observation time, and the measurement noise on damage detection performance. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric measurements in both standard and damaged conditions. A comparison of OCC algorithms, such as principal component analysis (PCA), kernel principal component analysis (KPCA), Gaussian mixture model (GMM) and one-class classifier neural network (OCCNN)^2 is performed, and their robustness to data shrinking is evaluated. In many cases, OCCNN^2 increases the performance with respect to classical anomaly detection techniques in terms of accuracy.

Dynamic Measurement of Portos Tomato Seedling Growth Using the Kinect 2.0 Sensor

Traditionally farmers monitor their crops employing their senses and experience. However, the human sensory system is inconsistent due to stress, health, and age. In this paper, we propose an agronomic application for monitoring the growth of Portos tomato seedlings using Kinect 2.0 to build a more accurate, cost-effective, and portable system. The proposed methodology classifies the tomato seedlings into four categories: The first corresponds to the seedling with normal growth at the time of germination; the second corresponds to germination that occurred days after; the third category entails exceedingly late germination where its growth will be outside of the estimated harvest time; the fourth category corresponds to seedlings that did not germinate. Typically, an expert performs this classification by analyzing ten percent of the randomly selected seedlings. In this work, we studied different methods of segmentation and classification where the Gaussian Mixture Model (GMM) and Decision Tree Classifier (DTC) showed the best performance in segmenting and classifying Portos tomato seedlings.

Multi-source fault diagnosis of chiller plant sensors based on an improved ensemble empirical mode decomposition Gaussian mixture model

As the core of the heating, ventilation, and air-conditioning (HVAC) control system, the sensor directly determines whether the HVAC control system is running normally. Poor natural and artificial environments can cause sensors to simultaneously fault, called multi-source sensor faults. Compared to the other methods, the Gaussian mixture model (GMM) has a better effect on high-dimensional data classification. In this paper, an improved ensemble empirical mode decomposition hard threshold denoising (EEMD-HTD) is first proposed to denoise the collected data and highlight each sensor’s fault characteristics. Secondly, the single-source and multi-source faults of the sensors are analyzed by the GMM at different points. The above fault diagnosis processes are collectively referred to as EEMD-HTD-GMM. Experimental results show that the EEMD-HTD-GMM has a higher diagnostic ability and distinguishes between the fault types that cause GMM classification confusion.

Stress Level Detection and Evaluation from Phonation and PPG Signals Recorded in an Open-Air MRI Device

This paper deals with two modalities for stress detection and evaluation—vowel phonation speech signal and photo-plethysmography (PPG) signal. The main measurement is carried out in four phases representing different stress conditions for the tested person. The first and last phases are realized in laboratory conditions. The PPG and phonation signals are recorded inside the magnetic resonance imaging scanner working with a weak magnetic field up to 0.2 T in a silent state and/or with a running scan sequence during the middle two phases. From the recorded phonation signal, different speech features are determined for statistical analysis and evaluation by the Gaussian mixture models (GMM) classifier. A database of affective sounds and two databases of emotional speech were used for GMM creation and training. The second part of the developed method gives comparison of results obtained from the statistical description of the sensed PPG wave together with the determined heart rate and Oliva–Roztocil index values. The fusion of results obtained from both modalities gives the final stress level. The performed experiments confirm our working assumption that a fusion of both types of analysis is usable for this task—the final stress level values give better results than the speech or PPG signals alone.