Mel Frequency Cepstral Coefficients Features Research Articles

An FMCW Radar-Based Intelligent System for Non-Contact Detection and Monitoring of Pneumonia Symptoms

Pneumonia is one of the most common contagious respiratory diseases, and one of its symptoms is shortness of breath. This symptom underscores the need for non-contact monitoring methods, which our paper addresses by proposing a strategy that uses Frequency-Modulated Continuous Wave (FMCW) radar to extract breathing waveforms and then classifies them with an eXtreme Gradient Boosting (XGBoost) model. The model performs well on our dataset, using stratified k-fold cross-validation and Mel-Frequency Cepstral Coefficients (MFCC) feature extraction. This intelligent system can correctly identify deep and deep-quick breathing patterns with 98% and 87.5% recall scores, respectively. Integrating FMCW and XGBoost offers a promising solution for early detection and real-time monitoring of pneumonia

Acoustic tunnel lining cavity detection using cepstral coefficients with optimized filter bank

Tunnels are an essential component of modern transportation infrastructure, and their structural health is critical to traffic safety, which can be seriously affected by tunnel lining cavities. In this paper, an acoustic-based detection approach for assessing the integrity of tunnel linings is studied. By tapping the tunnel lining surface, acoustic signals are sampled and analyzed using a novel feature parameter extraction algorithm-the energy-frequency cepstral coefficient, which uses wavelet packet decomposition to obtain energy distribution statistics in the frequency domain of the signal, and constructs a signal-dependent filter bank to achieve the cepstral coefficient extraction. Compared with the traditional Mel filter bank, this method can adaptively adjust the resolution of the filter bank according to the frequency characteristics of the classified samples. This allows for higher frequency resolution in regions where the energy distribution is concentrated. As a result, the extracted feature parameters achieve both dimensional compression and superior information retention. Experimental results show that the proposed energy-frequency cepstral coefficient feature outperforms the traditional Mel-frequency cepstral coefficient feature, resulting in a higher accuracy of tunnel lining detection. The convolutional neural network model achieves an accuracy of 99.2%, with a 78.9% reduction in error rate compared with the traditional Mel-frequency cepstral coefficient feature parameters. Additionally, a particle swarm optimization support vector machine model is trained to achieve an accuracy rate of 99.6% and an error rate reduction of 76.5%.

A novel concatenated 1D-CNN model for speech emotion recognition

Speech Emotion Recognition (SER) has conventionally been executed solely through acoustic data. Convolutional Neural Network (CNN) in Deep Learning (DL) is a cutting-edge technique that has been effectively used in diverse fields, such as analyzing speech emotion. This study utilizes the Mel Frequency Magnitude Coefficient (MFMC), a modified variant of the Mel Frequency Cepstral Coefficient (MFCC) feature, in conjunction with a one-dimensional (1D) CNN to increase the SER performance. The accuracy of SER is evaluated utilizing the proposed models. Model 1 refers to MFCC-1D-CNN, model 2 refers to MFMC-1D-CNN, and model 3 refers to the concatenated model of model 1 and model 2. The models were evaluated on four datasets comprising eight distinct emotions, specifically anger, happiness, sadness, boredom, fear, surprise, disgust, neutral, and calm. The proposed models, namely model 1, model 2, and model 3, were evaluated on the four datasets using coefficients of 12, 24, and 30. The Proposed Model 1 achieved accuracy rates for EMO-DB, EMOVO, SAVEE, and RAVDESS of 88.1%, 87.5%, 75%, and 97.7% for the respective coefficients. While the proposed model 2 achieved accuracy rates of 93.8%, 98.3%, 90.3%, and 97.5% for the same coefficients. However, the proposed concatenated model 3 outperformed the accuracy of 95.6%, 99.4%, 91.7%, and 98.1% for the four datasets EMO-DB, EMOVO, SAVEE, and RAVDESS compared to model 1 and model 2. Experimental results indicate that the proposed concatenated model 3 improves accuracy and prevents the model from overfitting.

Comparative study of respiratory sounds classification methods based on cepstral analysis and artificial neural networks

In this paper, we investigated and evaluated various machine learning-based approaches for automatically detecting wheezing sounds. We conducted a comprehensive comparison of these proposed systems, assessing their classification performance through metrics such as Sensitivity, Specificity, and Accuracy. The main approach to developing a machine learning-based system for classifying respiratory sounds involved the combination of a technique for extracting features from an unknown input sound with a classification method to determine its belonging class. The characterization techniques used in this study are based on the cepstral analysis, which was extensively employed in the automatic speech recognition field. While MFCC (Mel-Frequency Cepstral Coefficients) feature extraction methods are commonly used in respiratory sounds classification, our study introduces a novelty by employing GFCC (Gammatone-Frequency Cepstral Coefficients) and BFCC (Bark-Frequency Cepstral Coefficients) for this purpose. For the classification task, we employed two types of neural networks: the MLP (Multilayer Perceptron), a feedforward neural network, and a variant of the LSTM (Long Short-Term Memory) recurrent neural network called BiLSTM (Bidirectional LSTM). The proposed classification systems are evaluated using a database consisting of 497 wheezing segments and 915 normal respiratory segments, which are recorded from individuals diagnosticated with asthma and individuals without any respiratory issues, respectively. The highest classification performance was achieved by the BFCC-BiLSTM model, which demonstrated an exceptional accuracy rate of 99.8%.

An automatic method using MFCC features for sleep stage classification

Sleep stage classification is a necessary step for diagnosing sleep disorders. Generally, experts use traditional methods based on every 30 seconds (s) of the biological signals, such as electrooculograms (EOGs), electrocardiograms (ECGs), electromyograms (EMGs), and electroencephalograms (EEGs), to classify sleep stages. Recently, various state-of-the-art approaches based on a deep learning model have been demonstrated to have efficient and accurate outcomes in sleep stage classification. In this paper, a novel deep convolutional neural network (CNN) combined with a long short-time memory (LSTM) model is proposed for sleep scoring tasks. A key frequency domain feature named Mel-frequency Cepstral Coefficient (MFCC) is extracted from EEG and EMG signals. The proposed method can learn features from frequency domains on different bio-signal channels. It firstly extracts the MFCC features from multi-channel signals, and then inputs them to several convolutional layers and an LSTM layer. Secondly, the learned representations are fed to a fully connected layer and a softmax classifier for sleep stage classification. The experiments are conducted on two widely used sleep datasets, Sleep Heart Health Study (SHHS) and Vincent’s University Hospital/University College Dublin Sleep Apnoea (UCDDB) to test the effectiveness of the method. The results of this study indicate that the model can perform well in the classification of sleep stages using the features of the 2-dimensional (2D) MFCC feature. The advantage of using the feature is that it can be used to input a two-dimensional data stream, which can be used to retain information about each sleep stage. Using 2D data streams can reduce the time it takes to retrieve the data from the one-dimensional stream. Another advantage of this method is that it eliminates the need for deep layers, which can help improve the performance of the model. For instance, by reducing the number of layers, our seven layers of the model structure takes around 400 s to train and test 100 subjects in the SHHS1 dataset. Its best accuracy and Cohen’s kappa are 82.35% and 0.75 for the SHHS dataset, and 73.07% and 0.63 for the UCDDB dataset, respectively.

Machine learning-based infant crying interpretation.

Crying is an inevitable character trait that occurs throughout the growth of infants, under conditions where the caregiver may have difficulty interpreting the underlying cause of the cry. Crying can be treated as an audio signal that carries a message about the infant's state, such as discomfort, hunger, and sickness. The primary infant caregiver requires traditional ways of understanding these feelings. Failing to understand them correctly can cause severe problems. Several methods attempt to solve this problem; however, proper audio feature representation and classifiers are necessary for better results. This study uses time-, frequency-, and time-frequency-domain feature representations to gain in-depth information from the data. The time-domain features include zero-crossing rate (ZCR) and root mean square (RMS), the frequency-domain feature includes the Mel-spectrogram, and the time-frequency-domain feature includes Mel-frequency cepstral coefficients (MFCCs). Moreover, time-series imaging algorithms are applied to transform 20 MFCC features into images using different algorithms: Gramian angular difference fields, Gramian angular summation fields, Markov transition fields, recurrence plots, and RGB GAF. Then, these features are provided to different machine learning classifiers, such as decision tree, random forest, K nearest neighbors, and bagging. The use of MFCCs, ZCR, and RMS as features achieved high performance, outperforming state of the art (SOTA). Optimal parameters are found via the grid search method using 10-fold cross-validation. Our MFCC-based random forest (RF) classifier approach achieved an accuracy of 96.39%, outperforming SOTA, the scalogram-based shuffleNet classifier, which had an accuracy of 95.17%.

Data-centric artificial olfactory system based on the eigengraph

Recent studies of electronic nose system tend to waste significant amount of important data in odor identification. Until now, the sensitivity-oriented data composition has made it difficult to discover meaningful data to apply artificial intelligence in terms of in-depth analysis for odor attributes specifying the identities of gas molecules, ultimately resulting in hindering the advancement of the artificial olfactory technology. Here, we realize a data-centric approach to implement standardized artificial olfactory systems inspired by human olfactory mechanisms by formally defining and utilizing the concept of Eigengraph in electrochemisty. The implicit odor attributes of the eigengraphs were mathematically substantialized as the Fourier transform-based Mel-Frequency Cepstral Coefficient feature vectors. Their effectiveness and applicability in deep learning processes for gas classification have been clearly demonstrated through experiments on complex mixed gases and automobile exhaust gases. We suggest that our findings can be widely applied as source technologies to develop standardized artificial olfactory systems.

Developing children's ASR system under low-resource conditions using end-to-end architecture

The work presented in this paper aims at enhancing the performance of end-to-end (E2E) speech recognition task for children's speech under low resource conditions. For majority of the languages, there is hardly any speech data from child speakers. Furthermore, even the available children's speech corpora are limited in terms of the number of hours of data. On the other hand, large amounts of adults' speech data are freely available for research as well as commercial purposes. As a consequence, developing an effective E2E automatic speech recognition (ASR) system for children becomes a very challenging task. One may develop an ASR system using adults' speech and then use it to transcribe children's data, but this leads to very poor recognition rates due to the stark differences in the acoustic attributes of adults' and children's speech. In order to overcome these hurdles and to develop a robust children's ASR system employing E2E architecture, we have resorted to several out-of-domain and in-domain data augmentation techniques. For out-of-domain data augmentation, we have explicitly modified adults' speech to render it acoustically similar to that of children's speech before pooling into training. On the other hand, in the case of in-domain data augmentation, we have slightly modified the pitch and duration of children's speech in order to create more data capturing greater diversity. Data augmentation approaches helps in mitigating the ill-effects resulting from the scarcity of data from child domain to a certain extent. This, in turn, reduces the error rates by a large margin. In addition to data augmentation, we have also studied the efficacy of Gamma-tone frequency cepstral coefficients (GFCC) and frequency domain linear prediction (FDLP) technique along with the most commonly used Mel-frequency cepstral coefficients (MFCC) for front-end speech parameterization. Both MFCC as well as GFCC capture and model the spectral envelope of speech. On the other hand, application of linear prediction on the frequency domain representation of speech signal helps to effectively capture the temporal envelope during front-end feature extraction. Employing FDLP features that model the temporal envelope provides important cues for the perception and understanding of stop bursts and, at times, complete phonemes. This motivated us to perform a comparative experimental study of the effectiveness of the three aforementioned front-end acoustic features. In our experimental explorations, the use of proposed data augmentation in combination of FDLP features has shown a relative improvement in character error rate by 67.6% over the baseline system. The combination of data augmentation with MFCC or GFCC features is observed to result in lower recognition performances.

Significance of relative phase features for shouted and normal speech classification

Shouted and normal speech classification plays an important role in many speech-related applications. The existing works are often based on magnitude-based features and ignore phase-based features, which are directly related to magnitude information. In this paper, the importance of phase-based features is explored for the detection of shouted speech. The novel contributions of this work are as follows. (1) Three phase-based features, namely, relative phase (RP), linear prediction analysis estimated speech-based RP (LPAES-RP) and linear prediction residual-based RP (LPR-RP) features, are explored for shouted and normal speech classification. (2) We propose a new RP feature, called the glottal source-based RP (GRP) feature. The main idea of the proposed GRP feature is to exploit the difference between RP and LPAES-RP features to detect shouted speech. (3) A score combination of phase- and magnitude-based features is also employed to further improve the classification performance. The proposed feature and combination are evaluated using the shouted normal electroglottograph speech (SNE-Speech) corpus. The experimental findings show that the RP, LPAES-RP, and LPR-RP features provide promising results for the detection of shouted speech. We also find that the proposed GRP feature can provide better results than those of the standard mel-frequency cepstral coefficient (MFCC) feature. Moreover, compared to using individual features, the score combination of the MFCC and RP/LPAES-RP/LPR-RP/GRP features yields an improved detection performance. Performance analysis under noisy environments shows that the score combination of the MFCC and the RP/LPAES-RP/LPR-RP features gives more robust classification. These outcomes show the importance of RP features in distinguishing shouted speech from normal speech.

Comparative analysis of audio classification with MFCC and STFT features using machine learning techniques

In the era of automated and digitalized information, advanced computer applications deal with a major part of the data that comprises audio-related information. Advancements in technology have ushered in a new era where cutting-edge devices can deliver comprehensive insights into audio content, leveraging sophisticated algorithms such such as Mel Frequency Cepstral Coefficients (MFCCs) and Short-Time Fourier Transform (STFT) to extract and provide pertinent information. Our study helps in not only efficient audio file management and audio file retrievals but also plays a vital role in security, the robotics industry, and investigations. Beyond its industrial applications, our model exhibits remarkable versatility in the corporate sector, particularly in tasks like siren sound detection and more. Embracing this capability holds the promise of catalyzing the development of advanced automated systems, paving the way for increased efficiency and safety across various corporate domains. The primary aim of our experiment is to focus on creating highly efficient audio classification models that can be seamlessly automated and deployed within the industrial sector, addressing critical needs for enhanced productivity and performance. Despite the dynamic nature of environmental sounds and the presence of noises, our presented audio classification model comes out to be efficient and accurate. The novelty of our research work reclines to compare two different audio datasets having similar characteristics and revolves around classifying the audio signals into several categories using various machine learning techniques and extracting MFCCs and STFTs features from the audio signals. We have also tested the results after and before the noise removal for analyzing the effect of the noise on the results including the precision, recall, specificity, and F1-score. Our experiment shows that the ANN model outperforms the other six audio models with the accuracy of 91.41% and 91.27% on respective datasets.

Instrument Emotion Recognition from Polyphonic Instrumental Music using MFCC and CENS Features with Deep Neural Networks

The music digital data size is escalating in a manner where efficiently organizing and retrieving the data based on various parameters become tedious. Instrument music emotion recognition (IER) can be effectively utilized to recognize the emotion from instrumental music using machine learning methods. It has wide range of applications in music industry, psychology, medicine, and entertainment. Extracting or selecting suitable features that impact emotion recognition is one of the main tasks in IER. In this work, Mel Frequency Cepstral Coefficients (MFCC) and Chroma Energy Normalized Statistics (CENS) features have been employed to provide information to recognize the emotion pattern in instrumental music. These features have been trained using Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) which captures the high dimensional emotion patterns. A recognition rate of 89.7% and an equal error rate of 10.3% have been attained using CNN by combining the features. The experiment results depict that the recognition rate of each instrument-emotion class and the overall performance is improved by the score level fusion of the MFCC and CENS features compared to individual features.

Modulation and Energy Components for Speaker Recognition System

This paper presents a model use the modulation and energy components for speaker recognition application, that is mainly follows theshort-term scenario in speech signal processing, and also introduce a parameter combination that includes the instantaneous components and the energy parameters. This will describe the importance of short-term speech analysis in estimating the modulation parameters and the role of the instantaneous energy in estimating the speaker-dependent parameters. Simply, the short-term scenario is used to, first; avoid the silent and background noise speech portions that present in speech signals, and also to benefitfrom the stationary concept of the short-term processing in the speech signal. The energy components, on the other hand, are adopted purely in many speech parameterisation models, such as, linear predictive coding (LPC) and Mel-frequency cepstral coefficients (MFCCs). The main ideal of our mixture parameter (or MFCC/AM-FM model) is to determined the extent of these components to contribute together in extracting the parameters that are more related to the speaker more than anything else presented in the speech signal. We evaluated both models using the text-dependent and text-independent speech corpora. The accuracy results show that the frame-based AM-FM model achieve better performance comparing with the traditional structure of the AM-FM modulation model(the model presented in [1]). The MFCC/AM-FM parameters, on the other hand, perform much better, in terms of text-dependent, comparing with the AM-FM parameters and the MFCC parameters. In the case of the text-independent, however, the MFCC/AM-FM model provide better results than the MFCC features but less performance comparing to the AM-FM modulation parameters.

A Nature Inspired Optimization Algorithm for Parkinson's Disease Classification Through Speech Analysis

This research takes a unique approach to improve the accuracy of Parkinson's disease prediction through speech signal analysis by turning to the natural world for inspiration. This research proposes an optimization algorithm modelled on the hunting strategies of eagles. This approach begins with the extraction of Mel-frequency cepstral coefficients (MFCCs) from the audio signals. Subsequently, the proposed algorithm, inspired by the keen eyesight and hunting strategy of eagles, is employed to select the most relevant MFCC features. This optimization ensures a reduced dimensionality while maintaining the integrity and relevance of the feature set. Eagle's highly effective three-pronged approach to hunting - consisting of soaring (exploring), scanning (assessing), and swooping (taking action) - is the starting point for this method. To put this into practice, the proposed algorithm treats each potential combination of audio signal features as a 'solution', or 'prey'. The usefulness or 'fitness' of each solution is evaluated according to how accurately it can predict Parkinson's disease using a Support Vector Machine (SVM) classifier on a validation dataset. At the start, our algorithm generates an array of solutions. It then takes an 'eagle's eye view', assessing and refining these solutions in successive iterations. This mimics the eagle's hunt, where it identifies its target from a distance, before swooping in to capture it. By testing the proposed eagle-inspired algorithm on the Parkinson's Disease Voice Initiative (PDVI) dataset, it was found that it offers better accuracy in differentiating between those with Parkinson's and those without, compared to traditional methods. Specifically, using the proposed algorithm's selected features, the SVM model achieved an accuracy of 93.2% compared to an 88.5% accuracy when using all MFCC features. Furthermore, the computation time was reduced by 15% due to the dimensionality reduction introduced by the proposed algorithm. In conclusion, this study opens up a promising new direction in the early detection and treatment of Parkinson's disease, demonstrating the value of looking to nature for solving complex problems.

Unravelling stress levels in continuous speech through optimal feature selection and deep learning

Stress is a psychological or emotional strain that occurs due to adverse experiences in human life. This paper showcases the application of deep learning in detecting stress levels in continuous audio signals in the Distress Analysis Interview Corpus Wizard of Oz (DAIC-WOZ) database. The features that have been experimented with are Gammatone Frequency Cepstral Coefficients (GFCC), Log Filter Bank (Log-Filter Bank), Mel Frequency Cepstral Coefficients (MFCC), chroma, and Linear Predictive Coding (LPC). Five deep learning models were evaluated: Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Bidirectional LSTM (Bi-LSTM), k-fold CNN with the k value as 5, and a fusion model of CNN, LSTM, and attention. Upon evaluating the performance metrics of all the models, it is concluded that the k-fold CNN model with k as 5 performs well with continuous audio signals. The model has achieved an accuracy of 80% when it is trained on MFCC, GFCC, and Log-F Bank features which are observed to be the optimal features in the stress analysis.

AUTOMATIC ASSESSMENT OF COMMUNICATION SKILL IN NON-CONVENTIONAL INTERVIEW SETTINGS: A COMPARATIVE STUDY

This paper presents a comparative study of various technologies for predicting communication skills in interview candidates through non-conventional methods. The study explores three distinct modalities: video, audio, and essay-based assessments. For video analysis, facial expressions and eye movements are employed to gauge candidate confidence and confusion. In the audio domain, Mel-Frequency Cepstral Coefficients (MFCC) features are extracted from audio recordings and processed using deep learning algorithms to assess confidence levels. For essay-based evaluations, features extracted from textual content are fed into an XGBoost model to predict writing skills, with applicability to both written and verbal text. Existing technologies have achieved an accuracy range of 80-90% for essay-based predictions; however, our implementation of XGBoost has demonstrated improved performance with an accuracy between 95-96%. The results highlight the effectiveness of our approach in enhancing communication skill assessments through advanced machine learning and deep learning techniques, setting a new benchmark in the field.

Mel-frequency cepstral coefficients feature extracted voice recognition task using atomic switch Ag/Ag<sub>2</sub>S device-based time-delayed reservoir computing

Mel-frequency cepstral coefficients feature extracted voice recognition task using atomic switch Ag/Ag<sub>2</sub>S device-based time-delayed reservoir computing

MFCC feature extraction for COPD detection

Mel Frequency Cepstral Coefficients (MFCC) are acoustic features, which are based on human auditory perception, focus on spectral properties and capture relevant features of the audio signal related to vocal tract shape and energy in different frequency bands. The methodology for feature extraction involves several stages. First, the audio signals go through a preprocessing stage in which they are initially normalized, then background noise is reduced, and finally, they pass through a median filter. In the next block of MFCC feature extraction, the first task was to split the audio signal into small frames. Then, a hamming window is applied to smooth the edges of each frame and the short time Fourier transform of each frame is calculated. Next, Mel filters are applied, which adjusts the representation of the frequency spectrum to human auditory perception. Finally, cepstral coefficients are calculated from the frequency spectrum. The MFCC coefficients are then used as input features for classifiers and machine learning algorithms.

A deep learning model for depression detection based on MFCC and CNN generated spectrogram features

Depression is one of the leading forms of mental health issues encountered by individuals of diverse age groups today worldwide. Like any other mental health concerns, depression too poses diagnostic challenges for medical practitioners and clinical experts, given obvious social reservations and lack of awareness and acceptance in the society. Since long researchers have been looking for methods to identify symptoms of depression among individuals from their speech and responses, by utilizing automation systems and computers. In this paper, we propose an audio based depression detection method, which relies on neural networks for audio spectrogram based feature extraction as well as classification between speech/response patterns of depressed vs. non-depressed persons. We adopt a multi-modal approach in our work, by combining Mel-Frequency Cepstral Coefficients (MFCC) features, as well as Spectrogram features extracted from an audio file, by a novel CNN network. Our CNN model demonstrates optimized residual blocks and the “glorot uniform” kernel initializer. The proposed method’s performance is assessed in both multi-modal and multi-feature trials. We show our results on standard benchmark datasets DAIC-WOZ and MODMA, which provide repositories of questionnaire and patient responses, relevant in identification of depressive symptoms. We have also tested our model on standard emotion recognition audio dataset, RAVDESS. The proposed model achieves detection accuracy of over 90% in DAIC-WOZ and MODMA, and over 85% in RAVDESS, which is proven to surpass the present state-of-the-art.

Experimental studies for improving the performance of children's speaker verification system using short utterances

The task of developing an automatic speaker verification (ASV) system for children's speech is a formidable one due to a number of reasons. The dearth of domain-specific data is one among them. The challenge further intensifies with the introduction of short utterances of speech, a relatively unexplored domain in the case of children's ASV. Voice-based biometric systems suffers miserably when speech data, inadequate both in volume as well as in duration, is used either for enrollment or verification. To circumvent the issue arising due to data scarcity, the work in this paper extensively explores in-domain as well as various out-of-domain data augmentation techniques. A data augmentation approach is proposed that encompasses both in-domain and out-of-domain data augmentation techniques. The in-domain data augmentation approach, incorporates speed perturbation of children's speech. The out-of-domain data used are from adult speakers which are known to have acoustic attributes in stark contrast to child speakers. The acoustic characteristics of the adult speech data in this study are altered on two fronts namely speech waveform modification and feature-level modification, in order to modify the adult acoustic features and render it acoustically similar to children's speech prior to augmentation. While the speech waveform modification involves various signal processing techniques like prosody modification, formant modification and voice-conversion. The feature-level modification on the other hand involves Vocal-tract length normalization technique (VTLN) which explicitly models and compensates for the ill-effects of variations in vocal tract length by linearly warping the frequency axis of speech signals. The proposed data augmentation approach helps not only in increasing the amount of training data but also in effectively capturing the missing target attributes which helps in boosting the verification performance. A relative improvement of 48.01% in equal error rate (EER) with respect to the baseline system is a testimony of it. Furthermore, the conventionally used Mel-frequency cepstral coefficients (MFCC) are known to average out the higher-frequency components. Prior literary works have shown that a significant amount of relevant acoustic information is available in the higher-frequency region of the children's speech. Therefore, effective preservation of higher-frequency contents in children's speech is of paramount importance which must be appropriately tackled for the development of a reliable and robust children's ASV system. In this regard, frame-level concatenation of the MFCC features with the Inverse-Mel-frequency cepstral coefficient (IMFCC) features is undertaken. The feature concatenation of MFCC and IMFCC is carried out with the sole intention of effectively preserving the higher-frequency contents in the children's speech data. The low canonical correlation existing between the MFCC and the IMFCC feature vectors provides the necessary impetus to go with their feature fusion. The feature concatenation approach, when combined with proposed data augmentation, helps in further improvement of the verification performance. The experimental results testify our claims, wherein we have achieved an overall relative reduction of 50.15% for equal error rate.

Highly discriminative and adaptive feature extraction method based on NMF–MFCC for event recognition of Φ-OTDR

To enhance the capability of phase-sensitive optical time domain reflectometers (Φ-OTDR) to recognize disturbance events, an improved adaptive feature extraction method based on NMF–MFCC is proposed, which replaces the fixed filter bank used in the traditional method to extract the mel-frequency cepstral coefficient (MFCC) features by a spectral structure obtained from the Φ-OTDR signal spectrum using nonnegative matrix factorization (NMF). Three typical events on fences are set as recognition targets in our experiments, and the results show that the NMF–MFCC features have higher distinguishability, with the corresponding recognition accuracy reaching 98.47%, which is 7% higher than that using the traditional MFCC features.