Mel Frequency Cepstrum Research Articles

The phonetics of sociophonetics: Validating acoustic approaches to Spanish /s/

In many varieties of Spanish, syllable- and word-final /s/ is subject to a process of reduction from sibilance [s] through aspiration [h] to deletion. Sociolinguistic studies have traditionally used a three-way classification scheme on the basis of impressionistic coding; however, in the last decade, instrumental acoustic measurements have been favored. The present study examines several potential acoustic correlates of Spanish /s/ variants, including the often-used center of gravity, in their ability to faithfully represent the original perceptually-based observation of variation. The results indicate that many measurements can capture the contrast between sibilance [s] and aspiration [h]; however, fewer measurements (center of gravity after high-pass filter, skewness after high-pass filter, intensity without high-pass filter, zero crossing rate after high-pass filter, mel-frequency cepstrum coefficient 1 without high-pass filter) are also capable of detecting the contrast between aspiration [h] and deletion without making a priori assumptions about the appearance of non-zero variants. We propose that a combination of these measurements using Principal Component Analysis, which extracts the commonalities in the measurements, better represents the [s] > [h] > 0 cline than any one measurement by itself. We discuss the need for stricter evaluations of acoustic correlates of sociophonetic categories, especially regarding consonantal variation.

Squeak and Rattle recognition based on speaker recognition

Squeak and Rattle problems seriously affect the quality of the vehicles. In the adjustment stage of the automobile development process, engineers mainly rely on subjective evaluation to diagnose squeak and rattle noises, which is prone to misjudgment, missed judgment and time-consuming problems. There are gaps in relevant objective evaluation methods. This article proposed that speaker identification is used to identify squeak and rattle noises from the audio recorded in the test, to achieve objective and accurate results. In this paper, four kinds of squeak and rattle noise audios are used as samples, Mel Frequency Cepstrum Coefficient is extracted as feature vector to construct target Gaussian mixture models (GMM). The expectation, variance and correlation coefficient of the Gaussian mixture model can be calculated to describe the difference of multiple squeak and rattle noises. Using samples to judge accuracy rate, the results indicated that the acceptance accuracy rate reaches 100 %, and the rejection accuracy rate reaches more than 95 %.

Time-Delay-Neural-Network-Based Audio Feature Extractor for Ultra-Low Power Keyword Spotting

In this brief, we propose an audio feature extractor, based on a time delay neural network (TDNN), for ultra-low power keyword spotting (KWS). Conventionally, mel-frequency cepstrum coefficients (MFCCs) are widely used as features for KWS. However, an analog-to-digital converter (ADC) with high precision and high sampling rate is required for computing MFCCs, which consumes large amount of power. In our proposed feature extractor, on the other hand, input audio signals are band-pass filtered, and then the filtered signals are processed in TDNN in the analog domain, which can substitute a high precision ADC with simple clocked comparators. Simulation results show that the power dissipation of the proposed feature extractor can be reduced by 88% compared to the conventional MFCC feature extractor.

Research on Multifeature Intelligent Correction of Spoken English.

For a long time, college English teaching in many Chinese universities has focused on cultivating students' reading abilities while ignoring the cultivation of students' speaking abilities, leaving many non-English majors unable to communicate in English even after years of English study. This paper outlines the entire design and development process for an intelligent correction system for spoken English, with a focus on the methods for implementing the functions of spoken English examination, question bank management, and marking. A multifeature fusion of SE (sample entropy) and MFCC (Mel frequency cepstrum coefficient) based speech emotion recognition method is proposed. It denotes the rate at which the SE nonlinear dynamic system generates new data. It can be used to describe the dynamic fluctuation of speech signals in response to various emotions. To process SE and its statistics, as well as MFCC, and calculate the probability that they belong to one of six emotions, the support vector machine is used. The spoken English recognition algorithm described in this paper has obvious performance improvements in many indicators, according to the experimental evaluation.

RANCANG BANGUN MODEL PENGIDENTIFIKASI SUARA HURUF HIJAIYAH DENGAN METODE MEL FREQUENCY CEPSTRUM COEFFICIENT DAN CONVOLUTIONAL NEURAL NETWORK

Learning the Qur'an is a necessity for a Muslim, because the Qur'an acts as a guideand way of life. One thing that is learned in the Qur'an is how to pronounce Hijaiyah letters orMakharijul letters. In learning Makharijul Letters it takes an ustaz or accompanying teacher whois limited by distance and time. To overcome the limitations of distance and time, a learningapplication model is needed that can be accessed without the limitations of distance and thetime. This is studied aim to develop the Hijaiyah letter recognition model using the MelFrequency Cepstrum Coefficient (MFCC) and Convulational Neural Network (CNN) methods.Based validation results, the model built using the MFCC and CNN methods can identify theletters read with an accuracy of 49.1%. Then based of the test result with Confusion Matrix, thismodel have a precision value of 50%, recall is 53%, and F-Measure is 0.514.

Mel Frequency Cepstral Coefficient and its Applications: A Review

Feature extraction and representation has significant impact on the performance of any machine learning method. Mel Frequency Cepstrum Coefficient (MFCC) is designed to model features of audio signal and is widely used in various fields. This paper aims to review the applications that the MFCC is used for in addition to some issues that facing the MFCC computation and its impact on the model performance. These issues include the use of MFCC for non-acoustic signals, adopting the MFCC alone or combining it with other features, the use of time series versus global representation of the MFCC, following the standard form of the MFCC computation versus modifying its parameters, and supplying the traditional machine learning methods versus the deep learning methods.

Age-Based Automatic Voice Conversion Using Blood Relation for Voice Impaired

The present work presents a statistical method to translate human voices across age groups, based on commonalities in voices of blood relations. The age-translated voices have been naturalized extracting the blood relation features e.g., pitch, duration, energy, using Mel Frequency Cepstrum Coefficients (MFCC), for social compatibility of the voice-impaired. The system has been demonstrated using standard English and an Indian language. The voice samples for resynthesis were derived from 12 families, with member ages ranging from 8–80 years. The voice-age translation, performed using the Pitch synchronous overlap and add (PSOLA) approach, by modulation of extracted voice features, was validated by perception test. The translated and resynthesized voices were correlated using Linde, Buzo, Gray (LBG), and Kekre’s Fast Codebook generation (KFCG) algorithms. For translated voice targets, a strong (θ >~93% and θ >~96%) correlation was found with blood relatives, whereas, a weak (θ

Online Evaluation of Surface Hardness for Aluminum Alloy in LSP Using Modal Acoustic Emission

This article researches the online evaluation method of surface integrity for aluminum alloy in laser shock peening (LSP) based on multiple-sources modal acoustic emission (MAE). First, a new comprehensive index e.g., sub-surface microhardening rate (SHR) for 7075 aluminum alloy in LSP was constructed to characterize both the impact depth (<inline-formula> <tex-math notation="LaTeX">$150 \mu \text{m}$ </tex-math></inline-formula>) and its hardening rate. Then, a good linear and positive correlation between the SHR of 7075 Al alloy and the AE features in time domain and frequency domain was revealed based on the offline quantitative characterization of SHR and the antisymmetric zero (A0) mode separated from AE signal. In addition, the A0 mode-based Mel frequency cepstrum coefficient (AMFCC) was proposed to deeply extract the time-frequency features of MAE-A0 mode signal. Finally, the SHR classification model based on AMFCC and gradient boosting decision tree for 7075 Al alloy in LSP process was established and was carefully validated using the experimental data. The test results showed the highest mean accuracy of 96.24&#x0025; after the thorough comparison with other tree models and the traditional cepstrum methods. More importantly, deep analysis about the feature importance and their physical interpretation has been conducted from the perspective of non-linear effect inside the LSP material caused by dislocation multiplication as well as the increasing harmonic component of the MAE-A0 signal. This article can provide certain guidance for the development of LSP quality monitoring technology.

Retracted] Automatic Classification Method of Music Genres Based on Deep Belief Network and Sparse Representation

Aiming at the problems of poor classification effect, low accuracy, and long time in the current automatic classification methods of music genres, an automatic classification method of music genres based on deep belief network and sparse representation is proposed. The music signal is preprocessed by framing, pre‐emphasis, and windowing, and the characteristic parameters of the music signal are extracted by Mel frequency cepstrum coefficient analysis. The restricted Boltzmann machine is trained layer by layer to obtain the connection weights between layers of the depth belief network model. According to the output classification, the connection weights in the model are fine‐tuned by using the error back‐propagation algorithm. Based on the deep belief network model after fine‐tuning training, the structure of the music genre classification network model is designed. Combined with the classification algorithm of sparse representation, for the training samples of sparse representation music genre, the sparse solution is obtained by using the minimum norm, the sparse representation of test vector is calculated, the category of training samples is judged, and the automatic classification of music genre is realized. The experimental results show that the music genre automatic classification effect of the proposed method is better, the classification accuracy rate is higher, and the classification time can be effectively shortened.

Noise-robust text-dependent speaker identification using cochlear models.

One challenging issue in speaker identification (SID) is to achieve noise-robust performance. Humans can accurately identify speakers, even in noisy environments. We can leverage our knowledge of the function and anatomy of the human auditory pathway to design SID systems that achieve better noise-robust performance than conventional approaches. We propose a text-dependent SID system based on a real-time cochlear model called cascade of asymmetric resonators with fast-acting compression (CARFAC). We investigate the SID performance of CARFAC on signals corrupted by noise of various types and levels. We compare its performance with conventional auditory feature generators including mel-frequency cepstrum coefficients, frequency domain linear predictions, as well as another biologically inspired model called the auditory nerve model. We show that CARFAC outperforms other approaches when signals are corrupted by noise. Our results are consistent across datasets, types and levels of noise, different speaking speeds, and back-end classifiers. We show that the noise-robust SID performance of CARFAC is largely due to its nonlinear processing of auditory input signals. Presumably, the human auditory system achieves noise-robust performance via inherent nonlinearities as well.

What a deep song: The role of music features in perceived depth.

This study examines perceptions of music depth by exploring its relationships to different music features. First, a correlation analysis shows that the perceived depth of music is negatively correlated with valence and arousal and is also related to different music features, including tempo, Mel-frequency cepstrum coefficients, chromagrams, spectral centroids, spectral bandwidth, spectral contrast, spectral flatness, spectral roll-off, and tonal centroid features. Applying machine learning methods, we find that selected music features can predict perceptions of music depth, and a random forest regression (RFR) is found to perform best in this study. Finally, a feature importance analysis shows that the principal component of spectral contrast dominates the RFR-based music depth recognition model, showing that deep music usually has clear and narrow-band audio signals.

Fault detection of contactor using acoustic monitoring

Recently, condition monitoring methods using the sound of the machine have attracted attention. Since approaching high voltage equipment increases the risk of electrocuting, non-contact data acquisition is desirable. Most of the research targets of acoustic monitoring are rotating machines and it is not clear whether it is effective for machines that switch between two states, such as contactors and circuit breakers. In this work, several investigations have been carried out on the acoustic condition monitoring of contactor. The Mel-frequency cepstrum coefficients (MFCCs) were obtained from the sound data of the contactors under normal and simulated fault conditions. Support Vector Machine (SVM) was trained with MFCCs and found that it could detect and diagnose contactor faults with high accuracy.

Voice and Face Recognition for Web Browser Security

Abstract: This paper analysis about browser privacy multimodal authentication mechanisms. Face and voice recognition will be used as authentication methods in this process. The OpenCV library is used in the framework's face recognition section. It detects and recognizes faces from a database using basic eigen face recognition approaches. The MFCC (Mel Frequency Cepstrum Coefficients) and Gaussian Mixture Model are used to recognize voices. Following successful authentication, the cookies on the local hard disc are decrypted, allowing us access to the browser cookies. Initially, after a user registers, we will encrypt the browser cookies with AES, one of the most secure encryption methods available. keywords: MFCC, Gaussian Mixture Model, Browser cookies, authentication, AES, encryption, decryption, Open CV, Eigen.

Digital Audio Scene Recognition Method Based on Machine Learning Technology

Audio scene recognition is a task that enables devices to understand their environment through digital audio analysis. It belongs to a branch of the field of computer auditory scene. At present, this technology has been widely used in intelligent wearable devices, robot sensing services, and other application scenarios. In order to explore the applicability of machine learning technology in the field of digital audio scene recognition, an audio scene recognition method based on optimized audio processing and convolutional neural network is proposed. Firstly, different from the traditional audio feature extraction method using mel-frequency cepstrum coefficient, the proposed method uses binaural representation and harmonic percussive source separation method to optimize the original audio and extract the corresponding features, so that the system can make use of the spatial features of the scene and then improve the recognition accuracy. Then, an audio scene recognition system with two-layer convolution module is designed and implemented. In terms of network structure, we try to learn from the VGGNet structure in the field of image recognition to increase the network depth and improve the system flexibility. Experimental data analysis shows that compared with traditional machine learning methods, the proposed method can greatly improve the recognition accuracy of each scene and achieve better generalization effect on different data.

Bird Species Identification Using Spectrogram Based on Multi-Channel Fusion of DCNNs.

Deep convolutional neural networks (DCNNs) have achieved breakthrough performance on bird species identification using a spectrogram of bird vocalization. Aiming at the imbalance of the bird vocalization dataset, a single feature identification model (SFIM) with residual blocks and modified, weighted, cross-entropy function was proposed. To further improve the identification accuracy, two multi-channel fusion methods were built with three SFIMs. One of these fused the outputs of the feature extraction parts of three SFIMs (feature fusion mode), the other fused the outputs of the classifiers of three SFIMs (result fusion mode). The SFIMs were trained with three different kinds of spectrograms, which were calculated through short-time Fourier transform, mel-frequency cepstrum transform and chirplet transform, respectively. To overcome the shortage of the huge number of trainable model parameters, transfer learning was used in the multi-channel models. Using our own vocalization dataset as a sample set, it is found that the result fusion mode model outperforms the other proposed models, the best mean average precision (MAP) reaches 0.914. Choosing three durations of spectrograms, 100 ms, 300 ms and 500 ms for comparison, the results reveal that the 300 ms duration is the best for our own dataset. The duration is suggested to be determined based on the duration distribution of bird syllables. As for the performance with the training dataset of BirdCLEF2019, the highest classification mean average precision (cmAP) reached 0.135, which means the proposed model has certain generalization ability.

Deep learning-based classification of multichannel bio-signals using directedness transfer learning

The problem with processing of multivariate/multichannel signals lies in adapting of existing classifiers on data. Reformulating time-series data as visual clues and assigning visual patterns to different categories help the classification of time series in a wide range of applications. These series-to-image transformations have benefits including better noise robustness and more options regarding augmentation. They also provide the possibility of achieving discriminative features by employing transfer learning paradigm in cases dealing with highly small training datasets. In this respect, this work aimed to encode spectral-phase information into a bi-dimensional map. Transferring knowledge was done using bi-dimensional transformation capitalizing on the direction and propagation pattern of one channel influence on the others. EEG data from patients diagnosed with delirium (N = 15) recorded using a 10-channel BrainStatus device were used for this analysis. Considering leave-one-subject-out cross-validation, classification outcomes demonstrated that directedness transfer learning via Alexnet yields a promising performance showing 97.17% precision and outperforming other approaches. Comparison with nine different deep networks pretrained on ImageNet database was included. Directedness transfer learning resulted in precision of 95.29 ± 1.46 (µ ± σ)% among all networks. For further evaluation, directedness bi-dimensional transformation was also compared with six other 2D maps. Applying different networks resulted in average precision of (91.99 ± 2.23)% for polar-, (91.69 ± 1.57)% for correlation-, (90.46 ± 1.71)% for Spectrogram-, (87.82 ± 2.16)% for Wavelet-, (84.24 ± 1.72)% for Wigner-Ville- and (82.84 ± 2.46)% for Mel-frequency Cepstrum maps. To conclude, the proposed technique shows significant benefit in compressing spatio-spectral patterns of multichannel signals in just a unified visual representation.

Event Monitoring of Transformer Discharge Sounds based on Voiceprint

This paper investigates the operation inspection and anomaly diagnosis of transformers in substations, and carries out an application study of artificial intelligence-based sound recognition technology in transformer discharge diagnosis to improve the timeliness and diagnostic capability of intelligent monitoring of substation equipment operation. In this study, a sound parameterization technology in the field of sound recognition is used to implement automatic discharge sound detections. The sound samples are pre-processed and then Mel-frequency cepstrum coefficients (MFCCs) are extracted as features, which are used to train Gaussian mixture models (GMMs). Finally, the trained GMMs are used to detect discharge sounds in the place of transformers in substations. The test results demonstrate that the audio anomaly detection based on MFCCs and GMMs can be used to effectively recognize anomalous discharge in the high scenario of transformers.

Study on Gender Identification Based on Audio Recordings Using Gaussian Mixture Model and Mel Frequency Cepstrum Coefficient Technique

Speaker recognition is an ability to identify speaker’s characteristics based from spoken language. The purpose of this study is to identify gender of speakers based on audio recordings. The objective of this study is to evaluate the accuracy rate of this technique to differentiate the gender and also to determine the performance rate to classify even when using self-acquired recordings. Audio forensics uses voice recordings as part of evidence to solve cases. This study is mainly conducted to provide an easier technique to identify the unknown speaker characteristics in forensic field. This experiment is fulfilled by training the pattern classifier using gender dependent data. In order to train the model, a speech database is obtained from an online speech corpus comprising of both male and female speakers. During the testing phase, apart from the data from speech corpus, audio recordings of UTM students will too be used to determine the accuracy rate of this speaker identification experiment. As for the technique to run this experiment, Mel Frequency Cepstrum Coefficient (MFCC) algorithm is used to extract the features from speech data while Gaussian Mixture Model (GMM) is used to model the gender identifier. Noise removal was not used for any speech data in this experiment. Python software is used to extract using MFCC coefficients and model the behavior using GMM technique. Experiment results show that GMM-MFCC technique can identify gender regardless of language but with varying accuracy rate.

Automated sleep apnea detection in snoring signal using long short-term memory neural networks

Obstructive sleep apnea hypopnea syndrome (OSAHS) is a high incidence disease with serious hazard and potential danger. The polysomnography(PSG) has become the gold standard to diagnose OSAHS. However, the PSG is limited for household use because of its operational complexity, technical nature and high consumption.Currently, Microphone, as a non-contacting tensor, is an alternative method. And researchers devoted to analyze respiratory sound for detecting and evaluating OSAHS patients. In this paper, a classifier based on Long Short-Term Memory (LSTM) is proposed to identify the respiratory event-related snoring from simple snoring. Firstly, we collected the sleep sound of 33 patients and 10 normal people from the hospital. And 4780 abnormal snoring segments and 10,740 normal snoring segments were recorded by Mics. Then Mel-frequency cepstrum coefficients(MFCC), Mel Filter Banks (Fbanks), Short-time Energy and Linear Prediction Coefficient(LPC), representing the different characteristics of snoring, are extracted as characteristic features of snoring.At last, a multi-input model based on LSTM is designed, which can receive various audio features to synthesize information to identify snoring. Compared with single feature network processing, the use of multiple feature coefficients can identify the features of snoring at a fine-grained level. In the experiment, our method could classify respiratory event related snoring and normal snoring at accuracy 95.3%, and the accuracy of the three-category snore related to the severity of OSAHS can reach 81.6%. The recognition results can be used for the auxiliary diagnosis of OSAHS.

Real time Arabic speech recognition based on convolution neural network

The development of voice-activated devices to perform handless operation makes modern systems like consumer electronics, audio systems, automobile electronics, household appliances and televisions are smarter. Real-time speech recognition is a major element in such applications, and research is ongoing to improve the functioning of these systems in Arabic languages as well. This paper proposes an Arabic speech recognition system dependent on the Convolution neural networks. The speech recognition model using Mel-Frequency Cepstrum Coefficients (MFCCs) for feature extraction after being processed to reduce noise and remove silence from speech signals. The performance of the proposed system is evaluated on dataset content 12000 samples Arabic spoken digits and commands from various dialects. The proposed system is a better accuracy of recognition 100% during training phases, accuracy of validity 93, and 91.625 % accuracy of the CNN model testing.