Mel Frequency Cepstrum Coefficients Research Articles

A novel percussion-based approach for pipeline leakage detection with improved MobileNetV2

Pipelines are susceptible to oil and gas leaks during long-distance transportation due to factors such as damage from external forces and aging. However, existing pipeline leakage detection technologies that rely on physical inspections or sensors installed on pipelines are time-consuming and costly. In this paper, a percussion approach based on improved MobileNetV2 is proposed for pipeline leakage detection. Firstly, the influence of pipe leakage size on vibration characteristics was investigated by theoretical analysis and numerical simulation. Subsequently, experiments were conducted to assess the validity of the proposed method. The sounds produced by hammering the pipe under different damage conditions were recorded using a smartphone. The improved MobileNetV2 model was then used for classifying Mel spectrogram and Mel frequency cepstrum coefficient (MFCC) features extracted from the recorded sound signals. This model incorporates a multi-scale feature fusion module, which allows it to capture features at different scales and enhances its ability to differentiate between damage conditions. Experimental results show that using Mel spectrogram as input for the improved MobileNetV2 achieves a higher accuracy compared to using MFCC, with 100% accuracy for identifying leakage damage and 99.87% for classifying leak size. Compared to other methods, the improved MobileNetV2 exhibits superior classification performance while maintaining the lightweight characteristics of the original MobileNetV2. In conclusion, the improved model demonstrates significant enhancements in classification performance and operational efficiency, making it a promising approach for processing percussive signals.

9.1 µW keyword spotting processor based on optimized MFCC and small‐footprint TENet in 28‐nm CMOS

AbstractThis letter proposes a low‐power keyword spotting (KWS) architecture based on a modified temporal efficient neural network (TENet) and a simplified mel‐frequency cepstrum coefficient (MFCC) algorithm. The optimized MFCC algorithm reduces the computational load by 82% for multiplications and 66% for additions. An efficient hardware architecture and data flow for TENet have been designed, resulting in a 3.1× reduction in the operating cycle compared to similar work. The parameter count and computational load are reduced by 3.7× and 2.8×, respectively, and the accuracy reaches 95.36% for ten keywords in the Google Speech Command Dataset (GSCD). Operating at a frequency of 16 KHz for MFCC and 100 KHz for NN accelerator on a 28 nm process, the power consumption overhead is 9.1 µW.

Performance evaluation of lung sounds classification using deep learning under variable parameters

It is desired to apply deep learning models (DLMs) to assist physicians in distinguishing abnormal/normal lung sounds as quickly as possible. The performance of DLMs depends on feature-related and model-related parameters heavily. In this paper, the relationship between performance and feature-related parameters of a DLM, i.e., convolutional neural network (CNN) is analyzed through experiments. ICBHI 2017 is selected as the lung sounds dataset. The sensitivity analysis of classification performance of the DLM on three parameters, i.e., the length of lung sounds frame, overlap percentage (OP) of successive frames and feature type, is performed. An augmented and balanced dataset is acquired by the way of white noise addition, time stretching and pitch shifting. The spectrogram and mel frequency cepstrum coefficients of lung sounds are used as features to the CNN, respectively. The results of training and test show that there exists significant difference on performance among various parameter combinations. The parameter OP is performance sensitive. The higher OP, the better performance. It is concluded that for fixed sampling frequency 8 kHz, frame size 128, OP 75% and spectrogram feature is optimum under which the performance is relatively better and no extra computation or storage resources are required.

A Triplet Multimodel Transfer Learning Network for Speech Disorder Screening of Parkinson’s Disease

Deterioration in the quality of a person’s voice and speech is an early sign of Parkinson’s disease (PD). Although a number of computer-based methods have been invested to use patients’ speech for early diagnosis of Parkinson’s disease, they only focus on a fixed pronunciation test, such as the subjects’ monosyllabic pronunciation is analyzed to determine whether they have potential possibility of PD. Moreover, only using traditional speech analysis methods to extract single-view speech features cannot provide a comprehensive feature representation. This paper is dedicated to the study of various pronunciation tests for patients with PD, including the pronunciation of five monosyllabic vowels and a spontaneous dialogue. A triplet multimodel transfer learning network is designed and proposed for identifying subjects with PD in these two groups of tests. First, multisource data extract mel frequency cepstrum coefficient (MFCC) features of speech for preprocessing. Subsequently, a pretrained triplet model represents features from three dimensions as the upstream task of the transfer learning framework. Finally, the pretrained model is reconstructed as a novel model that integrates the triplet model, temporal model, and auxiliary layer as the downstream task, and weights are updated through fine-tuning to identify abnormal speech. Experimental results show that the highest PD detection rates in the two groups of tests are 99% and 90% , respectively, which outperform a large number of internationally popular pattern recognition algorithms and serve as a baseline for other academic researchers in this field.

Quran reciter identification using NASNetLarge

AbstractSpeaker identification has significant advantages for the field of human–computer interaction. Recently, many scholars have made contributions in this field and successfully created deep learning models for automatic speaker identification systems. However, most of the speech signal processing work is limited to English-only applications, despite numerous challenges with Arabic speech, particularly with the recitation of the Holy Quran, which is the Islamic holy book. In the light of these considerations, this study proposes a model for identifying the reciter of the Holy Quran using a dataset of 11,000 audio samples extracted from 20 Quran reciters. To enable feeding the audio samples' visual representation to the pre-trained models, the audio samples are converted from their original audio representation to visual representation using the Mel-Frequency Cepstrum Coefficients. Six pre-trained deep learning models are evaluated separately in the proposed model. The results from the test dataset reveal that the NASNetLarge model achieved the highest accuracy rate of 98.50% among the pre-trained models used in this study.

Automatic evaluation method for vehicle audio warning system using MFCC-polynomial hybrid feature

In the evaluation of vehicle audio warning system, there is no automatic method. Besides, due to the noise interference of in-vehicle environmental, the quantity limitation and only positive training samples, the accuracy of traditional template matching or identification methods for audio is low. To solve the above problems, an efficient, accurate, and automatic evaluation method is proposed for vehicle audio warning system. Firstly, logmmse-spectrum subtraction method is used to filter the dynamic noise and static noise of the evaluation audio acquired in the in-vehicle environment. Secondly, the end point detection based on short-time energy is used to obtain the effective audio segment after noise reduction, and the start time of the audio warning segment can be accurately obtained. Then, the Mel Frequency Cepstrum Coefficient (MFCC) feature and the polynomial fitting feature of each audio segment are extracted. The hybrid features are treated as the input of the Hidden Markov Model-Gaussian Mixture Model (GMM-HMM) based audio matching model. Finally, according to frame shift set by endpoint detection and the audio sampling frequency, the emitted time of matched audio warning can be calculated to support the evaluation of vehicle audio warning system. The experimental result shows that, with dynamic-static noise reduction and MFCC-polynomial hybrid feature, the average matching accuracy of the proposed method reaches 99.6% in the case of only five training samples.

Feature fusion method for pulmonary tuberculosis patient detection based on cough sound.

Since the COVID-19, cough sounds have been widely used for screening purposes. Intelligent analysis techniques have proven to be effective in detecting respiratory diseases. In 2021, there were up to 10 million TB-infected patients worldwide, with an annual growth rate of 4.5%. Most of the patients were from economically underdeveloped regions and countries. The PPD test, a common screening method in the community, has a sensitivity of as low as 77%. Although IGRA and Xpert MTB/RIF offer high specificity and sensitivity, their cost makes them less accessible. In this study, we proposed a feature fusion model-based cough sound classification method for primary TB screening in communities. Data were collected from hospitals using smart phones, including 230 cough sounds from 70 patients with TB and 226 cough sounds from 74 healthy subjects. We employed Bi-LSTM and Bi-GRU recurrent neural networks to analyze five traditional feature sets including the Mel frequency cepstrum coefficient (MFCC), zero-crossing rate (ZCR), short-time energy, root mean square, and chroma_cens. The incorporation of features extracted from the speech spectrogram by 2D convolution training into the Bi-LSTM model enhanced the classification results. With traditional futures, the best TB patient detection result was achieved with the Bi-LSTM model, with 93.99% accuracy, 93.93% specificity, and 92.39% sensitivity. When combined with a speech spectrogram, the classification results showed 96.33% accuracy, 94.99% specificity, and 98.13% sensitivity. Our findings underscore that traditional features and deep features have good complementarity when fused using Bi LSTM modelling, which outperforms existing PPD detection methods in terms of both efficiency and accuracy.

Machine Learning-Based Classification of Pulmonary Diseases through Real-Time Lung Sounds

    The study presents a computer-based automated system that employs machine learning to classify pulmonary diseases using lung sound data collected from hospitals. Denoising techniques, such as discrete wavelet transform and variational mode decomposition, are applied to enhance classifier performance. The system combines cepstral features, such as Mel-frequency cepstrum coefficients and gammatone frequency cepstral coefficients, for classification. Four machine learning classifiers, namely the decision tree, k-nearest neighbor, linear discriminant analysis, and random forest, are compared. Evaluation metrics such as accuracy, recall, specificity, and f1 score are employed. This study includes patients affected by chronic obstructive pulmonary disease, asthma, bronchiectasis, and healthy individuals. The results demonstrate that the random forest classifier outperforms the others, achieving an accuracy of 99.72% along with 100% recall, specificity, and f1 scores. The study suggests that the computer-based system serves as a decision-making tool for classifying pulmonary diseases, especially in resource-limited settings.

Pump Fault Detection Based on MFCC-MLCNN

 Detection of industrial water pumping systems is both a practical and important area of research in industrial production. The accuracy of fault detection is crucial because failure of fault detection can lead to pump damage and reduced productivity. In order to timely and accurately identify the working status of water supply pumps, a pump fault detection method based on Mel Frequency Cepstrum Coefficient with Migration Learning Convolutional Neural Networks (MFCC-MLCNN) is proposed. The sound signals of water pumps under different operating conditions are preprocessed to calculate their MFCC features as static features, and further processed to obtain the first-order difference MFCC features as well as the second-order difference MFCC features as dynamic features. In this study, the audio dataset of water pumps recorded with ambient noise in a real industrial environment is used. Usually, fault detection datasets are unbalanced because the amount of fault data is limited. A practical way to deal with this problem is to use deep migration learning, where high accuracy can be achieved with limited labeled data. In this study, a migration learning convolutional neural network is introduced to establish a pump fault diagnosis model. The experimental results show that the method proposed in this study can effectively recognize the working state of water supply pumps with a small amount of sample data and model parameters.

An Introduction to the Innovative Path and Exploration of the Teaching Reform of Ethnic Vocal Music

Abstract With the continuous development of literature and art, vocal education, as an important part of aesthetic education, has also been developed unprecedentedly. This paper constructs a music emotion recognition model based on deep learning and the Lstm network. The preprocessing of music is accomplished through the methods of pre-emphasis, frame-splitting, and windowing so as to improve the purity of the music signal. Using the Mel frequency cepstrum coefficient and cochlear frequency, the music analog signal is converted into frequency features so as to better distinguish the acoustic features and combined with Word2vec to realize the extraction of music emotion features. By comparing the Lstm music emotion recognition model with other models, the performance of this paper’s model is verified in terms of classification accuracy. To understand the role of its embodiment, the music recognition model is applied to the teaching of ethnic music. The results show that the average fitness of Lstm music recognition is 75%-90% with the increase of the number of evolutions, and the average fitness of Lstm objective function reaches the peak at the number of iterations of 40, with a fitness of 95%. Under the music recognition model, the students’ spectrum is elevated to 0db above the reference line, and the amplitude mostly floats in the interval in (-3, -15), and the teacher can formulate appropriate ethnic vocal music teaching for the students’ spectrum.

Mel Frequency Cepstral Coefficients (MFCC) Method and Multiple Adaline Neural Network Model for Speaker Identification

Speech recognition technology makes human contact with the computer more accessible. There are two phases in the speaker recognition process: capturing or extracting voice features and identifying the speaker's voice pattern based on the voice characteristics of each speaker. Speakers consist of men and women. Their voices are recorded and stored in a computer database. Mel Frequency Cepstrum Coefficients (MFCC) are used at the voice extraction stage with a characteristic coefficient of 13. MFCC is based on variations in the response of the human ear's critical range to frequencies (linear and logarithmic). The sound frame is converted to Mel frequency and processed with several triangular filters to get the cepstrum coefficient. Meanwhile, at the speech pattern recognition stage, the speaker uses an artificial neural network (ANN) Madaline model (many Adaline/ which is the plural form of Adaline) to compare the test sound characteristics. The training voice's features have been inputted as training data. The Madaline Neural Network training is BFGS Quasi-Newton Backpropagation with a goal parameter of 0,0001. The results obtained from the study prove that the Madaline model of artificial neural networks is not recommended for identification research. The results showed that the database's speech recognition rate reached 61% for ten tests. The test outside the database was rejected by only 14%, and 84% refused testing outside the database with different words from the training data. The results of this model can be used as a reference for creating an Android-based real-time system.

A Voice Disease Detection Method Based on MFCCs and Shallow CNN

The incidence rate of voice diseases is increasing year by year. The use of software for remote diagnosis is a technical development trend and has important practical value. Among voice diseases, common diseases that cause hoarseness include spasmodic dysphonia, vocal cord paralysis, vocal nodule, and vocal cord polyp. This paper presents a voice disease detection method that can be applied in a wide range of clinical. We cooperated with Xiangya Hospital of Central South University to collect voice samples from 352 different patients. The Mel Frequency Cepstrum Coefficient (MFCC) parameters are extracted as input features to describe the voice in the form of data. An innovative model combining MFCC parameters and single convolution layer CNN is proposed for fast calculation and classification. The highest accuracy we achieved was 92%, it is fully ahead of the original research results and internationally advanced. And we use advanced voice function assessment databases (AVFAD) to evaluate the generalization ability of the method we proposed, which achieved an accuracy rate of 98%. Experiments on clinical and standard datasets show that for the pathological detection of voice diseases, our method has greatly improved in accuracy and computational efficiency.

Audio signal based danger detection using signal processing and deep learning

Since there have been more and more incidents of women being harassed in the recent past, girls need to think twice before going out of their houses. Sometimes, they are not even safe in their house or workplace. These circumstance does not change for children of all genders who stay alone because of their working parents or for other reasons. Also, there is no such organized procedure to ensure safety and take women and children out of such violence and harassment. To address this problem, the authors of this paper developed an Android-based automated system to detect danger for women and children using audio from the surroundings. As the Android phone is available to everyone nowadays, they focused on using this device rather than developing a system on some external hardware. Different signal processing methods with deep learning techniques are used for this work. This work also addresses noise from the environment for any chaos and nullifies them using different noise reduction techniques such as Reduce Energy Noise, Reduce Mel Frequency Cepstrum Coefficient (MFCC) up Noise, Reduce Median Noise, Reduce Centroid Noise, Audio DeNoise, Noisereduce by Sainburf et al. & Butterworth high pass filter. The Noisereduce by Sainburg et al. along with the InceptionV3 model architecture turns out to be the best to classify audio with 95.51% accuracy. A new model called AudioViT is introduced. It uses a Visual Transformer and Residual Network to identify the audio signal. The Android application also takes necessary action when any unfavorable situation is detected. Android device users can use this application without any cost, which will pave the way to ensure the safety of women and children.

Forest Sound Event Detection with Convolutional Recurrent Neural Network-Long Short-Term Memory

Sound event detection tackles an audio environment's complex sound analysis and recognition problem. The process involves localizing and classifying sounds mainly to estimate the start point and end points of the separate sounds and describe each sound. Sound event detection capability relies on the type of sound. Although detecting sequences of distinct temporal sounds is straightforward, the situation becomes complex when the sound is multiple overlapping of much single audio. This situation usually occurs in the forest environment. Therefore, this aim of the paper is to propose a Convolution Recurrent Neural Network-Long Short-Term Memory algorithm to detect an audio signature of intruders in the forest environment. The audio is extracted in the Mel-frequency cepstrum coefficient and fed into the algorithm as an input. Six sound categories are chainsaw, machete, car, hatchet, ambiance, and bike. They were tested using several epochs, batch size, and filter of the layer in the model. The proposed model can achieve an accuracy of 98.52 percent in detecting the audio signature with a suitable parameter selection. In the future, additional types of audio signatures of intruders and further aspects of evaluation can be added to make the algorithm better at detecting intruders in the forest environment.

An ensemble-based framework for mispronunciation detection of Arabic phonemes

Determination of mispronunciations and ensuring feedback to users are maintained by computer-assisted language learning (CALL) systems. In this work, we introduce an ensemble model that defines the mispronunciation of Arabic phonemes and assists learning of Arabic, effectively. To the best of our knowledge, this is the very first attempt to determine the mispronunciations of Arabic phonemes employing ensemble learning techniques and conventional machine learning models, comprehensively. In order to observe the effect of feature extraction techniques, mel-frequency cepstrum coefficients (MFCC), and Mel-spectrogram are blended with each learning algorithm. To show the success of proposed model, 29 letters in the Arabic phonemes, 8 of which are hafiz, are voiced by a total of 11 different person. The amount of data set has been enhanced employing the methods of adding noise, time shifting, time stretching, pitch shifting. Extensive experiment results demonstrate that the utilization of voting classifier as an ensemble algorithm with Mel-spectrogram feature extraction technique exhibits remarkable classification result with 95.9% of accuracy.

Low Resource Language Analysis Using Deep Learning Algorithm for Gender Classification

Voice signals are the essential input source for applications based on human and computer interaction technology. Gender identification through voice signals is one of the most challenging tasks. For voice signal based analysis, deep learning algorithms provide an alternative to traditional and conventional algorithms for classification. To identify the gender through voice signals of female, male and ‘first-time’ transgender, the deep learning algorithm is used to improve the robustness of the identification model with the Mel Frequency Cepstrum Coefficients (MFCC) as a feature of the voice signals. This article presents the identification accuracy of gender with the help of recorded live voice signals. The voice samples of the third gender are recorded in the Hindi language. These Hindi language voice samples of transgender are very low resources and are unavailable at any recognized sources. The simulation results do not depend on the duration of the signals and are text independent. The recurrent neural network – Bidirectional Long Short-term Memory (RNN – BiLSTM) algorithm has been simulated on the recorded voice signals. The simulation outcome is compared with the earlier reported results in the literature. The gender-wise average accuracy of the proposed model is achieved as 91.44%, 94.94%, and 96.11% for males, females, and transgender, respectively, using voice signals. The identification accuracy of transgender is high in comparison to other genders. On the other hand, the average accuracy of the proposed model is obtained as 94.16%.

Human Age Estimation Through Audio Utilising MFCC and RNN

Age is one of human main attributes. Age is important factor to improve communication experience. Age estimation has been used in several applications to improve user experience. Therefore, an approach is needed to estimate the user age, one of which is through audio. In this study, Mel Frequency Cepstrum Coefficients (MFCC) and Recurrent Neural Network (RNN) will be used to estimate age through audio. MFCC is used to get features from audio data, while RNN is used to estimate age. Dataset used here was taken from corpus of user speech data on the Common Voice website. This study shows that MFCC and RNN methods are able to estimate human age through audio with highest accuracy obtained in SimpleRNN is 0.5647, and 0.7087 in LSTM.

Heart Rate Detection and Classification from Speech Spectral Features Using Machine Learning

Measurement of vital signs of the human body such as heart rate, blood pressure, body temperature and respiratory rate is an important part of diagnosing medical conditions and these are usually measured using medical equipment. In this paper, we propose to estimate an important vital sign – heart rate from speech signals using machine learning algorithms. Existing literature, observation and experience suggest the existence of a correlation between speech characteristics and physiological, psychological as well as emotional conditions. In this work, we estimate the heart rate of individuals by applying machine learning based regression algorithms to Mel frequency cepstrum coefficients, which represent speech features in the spectral domain as well as the temporal variation of spectral features. The estimated heart rate is compared with actual measurement made using a conventional medical device at the time of recording speech. We obtain estimation accuracy close to 94% between the estimated and actual measured heart rate values. Binary classification of heart rate as ‘normal’ or ‘abnormal’ is also achieved with 100% accuracy. A comparison of machine learning algorithms in terms of heart rate estimation and classification accuracy is also presented. Heart rate measurement using speech has applications in remote monitoring of patients, professional athletes and can facilitate telemedicine.

Exploring the ability of vocal biomarkers in distinguishing depression from bipolar disorder, schizophrenia, and healthy controls.

Vocal features have been exploited to distinguish depression from healthy controls. While there have been some claims for success, the degree to which changes in vocal features are specific to depression has not been systematically studied. Hence, we examined the performances of vocal features in differentiating depression from bipolar disorder (BD), schizophrenia and healthy controls, as well as pairwise classifications for the three disorders. We sampled 32 bipolar disorder patients, 106 depression patients, 114 healthy controls, and 20 schizophrenia patients. We extracted i-vectors from Mel-frequency cepstrum coefficients (MFCCs), and built logistic regression models with ridge regularization and 5-fold cross-validation on the training set, then applied models to the test set. There were seven classification tasks: any disorder versus healthy controls; depression versus healthy controls; BD versus healthy controls; schizophrenia versus healthy controls; depression versus BD; depression versus schizophrenia; BD versus schizophrenia. The area under curve (AUC) score for classifying depression and bipolar disorder was 0.5 (F-score = 0.44). For other comparisons, the AUC scores ranged from 0.75 to 0.92, and the F-scores ranged from 0.73 to 0.91. The model performance (AUC) of classifying depression and bipolar disorder was significantly worse than that of classifying bipolar disorder and schizophrenia (corrected p < 0.05). While there were no significant differences in the remaining pairwise comparisons of the 7 classification tasks. Vocal features showed discriminatory potential in classifying depression and the healthy controls, as well as between depression and other mental disorders. Future research should systematically examine the mechanisms of voice features in distinguishing depression with other mental disorders and develop more sophisticated machine learning models so that voice can assist clinical diagnosis better.

The structural tuning of the convolutional neural network forspeaker identification in mel frequency cepstrumcoefficients space

As a result of the literature analysis, the main methods for speaker identification from speech signals were defined. These are statistical methods based on Gaussian mixture model and a universal background model, as well as neural network methods, in particular, using convolutional or Siamese neural networks. The main characteristics of these methods are the recognition performance, a number of parameters, and the training time. High recognition performance is achieved by using convolutional neural networks, but a number of parameters of these networks are much higher than for statistical methods, although lower than for Siamese neural networks. A large number of parameters require a large training set, which is not always available for the researcher. In addition, despite the effectiveness of convolutional neural networks, model size and inference efficiency remain important fordevices with a limited source of computing power, such as peripheral or mobile devices. Therefore, the aspects of tuning of the structure of existing convolutional neural networks are relevant for research. In this work, we have performed a structural tuning of an existing convolutional neural network based on the VGGNet architecture for speaker identification in the space of mel frequency cepstrum coefficients. The aim of the work was to reduce the number of neural network parameters and, as a result, to reduce the network training time, provided that the recognition performance is sufficient (the correct recognition is above 95 %). The neural network proposed as a result of structural tuning has fewer layers than the architecture of the basic neural network. Insteadof the ReLU activation function, the related Leaky ReLU function with a parameter of 0.1 was used. The number of filters and the size of kernels in convolutional layers are changed. The size of kernels for the max pooling layer has been increased. It is proposedto use the averaging of the results of each convolution to input a two-dimensional convolution results to a fully connected layer with the Softmax activation function. The performed experiment showed that the number of parameters of the proposed neural network is 29 % less than the number of parameters of the basic neural network, provided that the speaker recognition performance is almost the same. In addition, the training time of the proposed and basic neural networks was evaluated on five datasets of audio recordings corresponding to different numbers of speakers. The training time of the proposed network was reduced by 10-39 % compared to the basic neural network. The results of the research show the advisability of the structural tuning of the convolutional neural network for devices with a limited source of computing, namely, peripheral or mobile devices.