Zero-crossing Rate Research Articles

Analisis Fitur dan Convolutional Neural Network pada Pengenalan Aksen Ucapan

Each country has its characteristics and culture, one of these characteristics is the accent of speech. By listening to someone's accent, we can identify the country of origin of the speaker. Research on accent recognition includes Automatic Speech Recognition (ASR) Technology which is currently being developed, an example of ASR technology, namely Virtual Assistant, the development of this research can be more intelligent Virtual Assistant because it can provide an accent from a speaker. In this study, the authors tried to classify accents from various countries (5 classes), namely English, Spanish, Mandarin, French and Arabic. The dataset used in this study consists of English 627 audio, Spanish 220 audio, Mandarin 132 audio, French 80 audio, and Arabic 172 audio, where all sentences are the same sentence in English. In this study, the audio features used are Mel - Frequency Cepstral Coefficients (MFCC), Zero Crossing Rate (ZCR), and Energy (in librosa it is called RMS). Audio feature extraction generates an array of each audio, the result of audio feature extraction will be the Convolutional Neural Network (CNN) Method input for classifying the accent. This research resulted in 51.30% accuracy for the MFCC feature, 48.05% for the ZCR feature, and 51.95% for the Energy feature. The Energy feature gets good accuracy, followed by the MFCC and ZCR features.

Quality control of microseismic P-phase arrival picks in coal mine based on machine learning

Microseismic events generally contain strong noise-polluting and unobvious P-phase oscillating channel waveforms. The automatic P-phase arrival picking accuracy of these channel waveforms tends to be low, or even are false. Currently, unusable P-picks are not screened out automatically before geophysics inversions in most microseismic data processing software. Therefore, manual interventions are needed to remove or correct the unusable P-picks. However, rapidly increasing monitoring data causes manual handling to be time-consuming and lagging. Supervised machine learning (ML) is applied to distinguish useable and unusable P-picks automatically. Big data analysis revealed that the waveform features, including signal-to-noise ratio, signal-to-noise variance ratio, P-wave starting-up slope, and peak amplitude have impact on P-pick accuracy. In contrast, the effect of the short-time zero-crossing rate on the P-pick accuracy is not as obvious. Five P-pick quality control models were trained based on traditional machine learning approaches, including discriminant analysis, logistic regression, k-nearest neighbor, support vector machine, and Naive Bayes classifier. For these five models, the input data are P-pick labels and waveform features. In addition, another P-pick quality control model was trained based on convolutional neural network. While, the input data are P-pick images and labels. The training sets used in all six machine learning models are uniform. The testing experiments with uniform testing set show that the support vector machine generated best the performance among traditional machine learning approaches, with 82.81% accuracy. However, the convolutional neural network model generated outstanding performance in recognizing P-pick, with 91.71% accuracy. The automatic P-pick quality control method proposed in this study can facilitate the precision and efficiency of the automatic processing of microseismic signals.

Objective Evaluation of Therapeutic Effects of ADHD Medication by Analyzing Movements Using a Smart Chair with Piezoelectric Material

Attention-deficit hyperactivity disorder (ADHD) is the most common neuropsychiatric disorder in schoolchildren. Several methods are available to evaluate ADHD therapeutic effects, including the Swanson, Nolan, and Pelham (SNAP) questionnaire, the Vanderbilt ADHD Diagnostic Rating Scale, and the visual analog scale. However, these scales are subjective. In this study, a piezoelectric material was applied to a medical chair to objectively evaluate the therapeutic effect of ADHD medication before and after treatment. A total of 22 patients (18 boys and 4 girls) with ADHD were enrolled. During the appointment, the patients’ movements were recorded by the piezoelectric material before being analyzed. The variance, zero-crossing rate, and high energy rate of movements were used to analyze the signal in this study. The results showed the variance, zero-crossing rate, and high energy rate in patients with ADHD all decreased significantly after 1 month of methylphenidate use. Although the hyperactivity subscales of SNAP obtained from parents and teachers demonstrated significant decreases after 1 month of medication, the reduction rate of the three aforementioned measurements decreased more than hyperactivity subscales. This suggests that the use of a smart chair equipped with a piezoelectric material is an objective and useful method for evaluating the therapeutic effects of ADHD medication.

Towards Characterizing and Developing Formation and Migration Cues in Seafloor Sand Waves on Topology, Morphology, Evolution from High-Resolution Mapping via Side-Scan Sonar in Autonomous Underwater Vehicles.

Sand waves constitute ubiquitous geomorphology distribution in the ocean. In this paper, we quantitatively investigate the sand wave variation of topology, morphology, and evolution from the high-resolution mapping of a side scan sonar (SSS) in an Autonomous Underwater Vehicle (AUV), in favor of online sequential Extreme Learning Machine (OS-ELM). We utilize echo intensity directly derived from SSS to help accelerate detection and localization, denote a collection of Gaussian-type morphological templates, with one integrated matching criterion for similarity assessment, discuss the envelope demodulation, zero-crossing rate (ZCR), cross-correlation statistically, and estimate the specific morphological parameters. It is demonstrated that the sand wave detection rate could reach up to 95.61% averagely, comparable to deep learning such as MobileNet, but at a much higher speed, with the average test time of 0.0018 s, which is particularly superior for sand waves at smaller scales. The calculation of morphological parameters primarily infer a wave length range and composition ratio in all types of sand waves, implying the possible dominant direction of hydrodynamics. The proposed scheme permits to delicately and adaptively explore the submarine geomorphology of sand waves with online computation strategies and symmetrically integrate evidence of its spatio-temporal responses during formation and migration.

Effects of Continuous Positive Airway Pressure on Sleep EEG Characteristics in Patients with Primary Central Sleep Apnea Syndrome.

This study aimed to investigate the effects of continuous positive airway pressure (CPAP) on the electroencephalographic (EEG) characteristics of patients with primary central sleep apnea syndrome (CSAS). Nine patients with primary CSAS were enrolled in this study. The raw sleep EEG data were analyzed based on two main factors: fractal dimension (FD) and zero-crossing rate of detrended FD. Additionally, conventional EEG spectral analysis in the delta, theta, alpha, and beta bands was conducted using a fast Fourier transform. The FD in patients with primary CSAS who underwent CPAP treatment was significantly decreased during nonrapid eye movement (NREM) sleep but increased during rapid eye movement (REM) sleep (p < 0.05). Regarding the EEG spectral analysis, the alpha power increased, while the delta/alpha ratio decreased during REM sleep in patients with CSAS (p < 0.05). In conclusion, CPAP treatment can reduce FD in NREM sleep and increase FD during REM sleep in patients with primary CSAS. FD may be used as a new biomarker of EEG stability and improvement in brain function after CPAP treatment for primary CSAS.

Short-term feeding behaviour sound classification method for sheep using LSTM networks

A deep learning approach using long-short term memory (LSTM) networks was implemented in this study to classify the sound of short-term feeding behaviour of sheep, including biting, chewing, bolus regurgitation, and rumination chewing. The original acoustic signal was split into sound episodes using an endpoint detection method, where the thresholds of short-term energy and average zero-crossing rate were utilized. A discrete wavelet transform (DWT), Mel-frequency cepstral, and principal-component analysis (PCA) were integrated to extract the dimensionally reduced DWT based Mel-frequency cepstral coefficients (denoted by PW_MFCC) for each sound episode. Then, LSTM networks were employed to train classifiers for sound episode category classification. The performances of the LSTM classifiers with original Mel-frequency cepstral coefficients (MFCC), DWT based MFCC (denoted by W_MFCC), and PW_MFCC as the input feature coefficients were compared. Comparison results demonstrated that the introduction of DWT improved the classifier performance effectively, and PCA reduced the computational overhead without degrading classifier performance. The overall accuracy and comprehensive F1-score of the PW_MFCC based LSTM classifier were 94.97% and 97.41%, respectively. The classifier established in this study provided a foundation for an automatic identification system for sick sheep with abnormal feeding and rumination behaviour pattern. Keywords: sheep behaviour, short-term feeding behaviour, acoustic analysis, Mel-frequency cepstral coefficients, long-short term memory networks DOI: 10.25165/j.ijabe.20211402.6081 Citation: Duan G H, Zhang S F, Lu M Z, Okinda C, Shen M X, Norton T. Short-term feeding behaviour sound classification method for sheep using LSTM networks. Int J Agric & Biol Eng, 2021; 14(2): 43–54.

Feature extraction of noise signal in motorcycle by Fast Fourier Transform

This research aimed to study the feature of noise signal from various exhausts in motorcycles by the Fast Fourier Transform (FFT). The noise signal was examined into three different features; energy, entropy, and zero-crossing rate. The energy and entropy defined the noise came from an exhaust or non-exhaust, while the zero-crossing rate distinguished normal and racing-exhaust. The results showed that the energy of an exhaust noise was always in the range of 2.4 to 2.8 joule. The entropy of exhausts racing was observed in the range of 0.00 to 0.05 joule. When the acceleration pedal was on, the entropy was become higher. The zero-crossing rate of racing exhaust was in the range of 0.00 to 0.05, the non-racing exhaust was in the range 0.05 to 0.1.

Speech emotion recognition based on SVM and KNN classifications fusion

Recognizing the sense of speech is one of the most active research topics in speech processing and in human-computer interaction programs. Despite a wide range of studies in this scope, there is still a long gap among the natural feelings of humans and the perception of the computer. In general, a sensory recognition system from speech can be divided into three main sections: attribute extraction, feature selection, and classification. In this paper, features of fundamental frequency (FEZ) (F0), energy (E), zero-crossing rate (ZCR), fourier parameter (FP), and various combinations of them are extracted from the data vector, Then, the principal component analysis (PCA) algorithm is used to reduce the number of features. To evaluate the system performance. The fusion of each emotional state will be performed later using support vector machine (SVM), K-nearest neighbor (KNN), In terms of comparison, similar experiments have been performed on the emotional speech of the German language, English language, and significant results were obtained by these comparisons.

A stacked auto-encoder with scaled conjugate gradient algorithm for Malayalam ASR

Automatic speech recognition (ASR) is entitled to automate natural speech perception and the processing mechanism through analysis in the linguistic and acoustic features of the speech signal. ASR for children is highly challenging due to their developing physical aspects and rapidly changing articulation features. Therefore, ASR for children is still at its infant level. In this work, a stacked multilayer auto-encoder (AE) network is designed for ASR of the Malayalam vowel, articulated by children in the age group of five to ten. The proposed network structured with an unsupervised pre-training followed by supervised training. The pre-training coupled with two layers of sparse auto-encoders and scaled conjugate gradient (SCG) algorithm used for back-propagation. The auto-encoders are used to pre-train the network in an unsupervised (self- supervised) manner with 40,500 features that include Mel frequency cepstral coefficients (MFCC) and its derivatives, spectrogram formants and zero crossing rate (ZCR). In the softmax layer, the pre-trained network retrained in a supervised manner with bottleneck features. Fine-tuning has been applied in the trained network to enhance its performance. The unsupervised and supervised layers are stacked together to form a comprehensive network. The designed network has shown an average accuracy of 97% in training and 89.5% accuracy in the test data-set.

A Proposed Speaker Recognition Method Based on Long-Term Voice Features and Fuzzy Logic

Speaker recognition depends on specific predefined steps. The most important steps are feature extraction and features matching. In addition, the category of the speaker voice features has an impact on the recognition process. The proposed speaker recognition makes use of biometric (voice) attributes to recognize the identity of the speaker. The long-term features were used such that maximum frequency, pitch and zero crossing rate (ZCR). In features matching step, the fuzzy inner product was used between feature vectors to compute the matching value between a claimed speaker voice utterance and test voice utterances. The experiments implemented using (ELSDSR) data set. These experiments showed that the recognition accuracy is 100% when using text dependent speaker recognition.

A novel audio watermarking scheme based on fuzzy inference system in DCT domain

Digital watermarking technology provides an effective potential solution to copyright protection and authentication of digital media. In this article, a new design of audio watermarking is introduced which is able to make compromise between transparency, robustness, and capacity by means of the synergy between fuzzy inference system, Singular Value Decomposition (SVD), and Fibonacci’s sequence in Discrete Cosine Transform (DCT). In the embedding phase, the proposed method, first of all, finds proper segments for inserting watermark data by helping fuzzy inference system with energy, zero-crossing rate (ZCR), and music edge features. Then, the watermark bits are embedded in the suitable segments based on collaboration of SVD technique and Fibonacci sequence in DCT domain. The proposed watermark extraction is performed in blind manner. The results on five Blues, Electronic, Classic, Jazz and Rock audio files show that the proposed method has high transparency (in average SNR = 49.80 dB) with payload of 598.34 bps. Moreover, robustness tests against Stirmark attacks show that the average of error rate is 1.3644, which means the proposed scheme has high stability in the digital signal processing attacks.

A multi-time scale vibration surveillance system for third-party threats on urban pipeline

Third-party interference caused by construction activities have seriously jeopardized the security of underground pipelines. Following the process of signal collection—feature extraction and selection—multi-time scale identifying— combining results by voting, this paper proposes a multi-time scale surveillance system for interference prevention of thirdparty threats on the nearby pipeline by using ground vibration monitors. The system focuses on the two major urban construction activities induced by excavator breaking hammers and road cutters, and presents excellent performance under the noise of traffic and pedestrian. Three features including the short-time zero-crossing rate, subset differential parameter and the Mel frequency cepstrum coefficients are selected by the analysis of the maximal information coefficient and feature importance for identifying the patterns of different third-party activities. The crucial part of the surveillance system consists of the two random forest-based classifiers trained by 0.5 s samples and 8 s samples respectively, and the alarm depends on the voting of the two classifiers, which brings the perspectives on different time scales for decision making. In the test, 96.14% of the threat vibration signals can be detected, while only 0.45% of the environmental noise signals cause false alarms.

Detection of Bryde's whale short pulse calls using time domain features with hidden Markov models

Passive acoustic monitoring (PAM) is generally usedto extract acoustic signals produced by cetaceans. However, the large data volume from the PAM process is better analyzed using an automated technique such as the hidden Markovmodels (HMM). In this paper, the HMM is used as a detection and classification technique due to its robustness and low time complexity. Nonetheless, certain parameters, such as the choice of features to be extracted from the signal, the frame duration, and the number of states affect the performance of the model. Theresults show that HMM exhibits best performances as the number of states increases with short frame duration. However, increasing the number of states creates more computational complexity in the model. The inshore Bryde's whales produce short pulse calls with distinct signal features, which are observable in the time-domain. Hence, a time-domain feature vector is utilized to reduce the complexity of the HMM. Simulation results also show that average power as a time-domain feature vector provides the best performance compared to other feature vectors for detecting the short pulse call of inshore Bryde's whales based on the HMM technique. More so, the extracted features such as the average power, mean, and zero-crossing rate, are combined to form a single 3-dimensional vector (PaMZ). The PaMZ-HMM shows improved performance and reduced complexity over existing feature extraction techniques such as Mel-scale frequency cepstral coefficients (MFCC) and linear predictive coding (LPC). Thus, making the PaMZ-HMM suitable for real-time detection.

An Interferometric Optical Fiber Perimeter Security System Based on Multi-Domain Feature Fusion and SVM

In this paper, an interferometric optical fiber perimeter security system based on multi-domain feature fusion and support vector machine (SVM) is reported. To improve the intrusion event classification accuracy and reduce the overall cost, advanced optical fiber sensor technique and data analysis method are synthesized to build the system. For the optical component, the Michelson interferometer sensor structure and the rectangular-pulse binary phase modulation method are adopted to reduce the system complexity and cost. For the electronic data analysis component, mixed domains (including time, frequency and wavelet domains) features are selected and the SVM is adopted to classify the intrusion event type. Specifically, the zero-crossing rate (ZCR) is chosen as the feature element in the time domain, whose threshold is determined by analyzing the environmental noise. The variance of the power spectral density (PSD) of the phase signal in four frequency bands are selected as the feature element in the frequency domain. The energy of the phase signal by using the wavelet packet decomposition in four full frequency bands is chosen as the feature element in the wavelet domain. All the acquired features corresponding to various event conditions are combined into the feature vector data set, which are trained and predicted by the SVM. In the field experiment test, the proposed identification scheme was used to identify and classify five kinds of events, including non-intrusion, climbing, shaking, iron bar knocking and fiber cable shearing. The achieved average classification accuracy reached 94.4%.

Automatic recognition of uterine contractions with electrohysterogram signals based on the zero-crossing rate

Uterine contraction (UC) is an essential clinical indicator in the progress of labour and delivery. Electrohysterogram (EHG) signals recorded on the abdomen of pregnant women reflect the uterine electrical activity. This study proposes a novel algorithm for automatic recognition of UCs with EHG signals to improve the accuracy of detecting UCs. EHG signals by electrodes, the tension of the abdominal wall by tocodynamometry (TOCO) and maternal perception were recorded simultaneously in 54 pregnant women. The zero-crossing rate (ZCR) of the EHG signal and its power were calculated to modulate the raw EHG signal and highlight the EHG bursts. Then the envelope was extracted from the modulated EHG for UC recognition. Besides, UC was also detected by the conventional TOCO signal. Taking maternal perception as a reference, the UCs recognized by EHG and TOCO were evaluated with the sensitivity, positive predictive value (PPV), and UC parameters. The results show that the sensitivity and PPV are 87.8% and 93.18% for EHG, and 84.04% and 90.89% for TOCO. EHG detected a larger number of UCs than TOCO, which is closer to maternal perception. The duration and frequency of UC obtained from EHG and TOCO were not significantly different (p > 0.05). In conclusion, the proposed UC recognition algorithm has high accuracy and simple calculation which could be used for real-time analysis of EHG signals and long-term monitoring of UCs.

Speech file compression by eliminating unvoiced/silence components

The main objective of this study is to have the noise component of a speech signal eliminated and compressed by storing the locations and durations of silence regions. The separation between voiced, unvoiced, and silence regions is done by using the Short-Time Energy (STE) and Zero Crossing Rate (ZCR) methodologies. All operations in this study have been performed by using the User Interface (UI) developed on MATLAB®. These operations include voice recording, playing the recording, eliminating the unwanted regions, playing the modified recording, saving original and compressed files, and loading the recording compressed.

Inter classifier comparison to detect voice pathologies.

Voice pathologies are irregular vibrations produced due to vocal folds and various factors malfunctioning. In medical science, novel machine learning algorithms are applied to construct a system to identify disorders that occur invoice. This study aims to extract the features from the audio signals of four chosen diseases from the SVD dataset, such as laryngitis, cyst, non-fluency syndrome, and dysphonia, and then compare the four results of machine learning algorithms, i.e., SVM, Naïve Byes, decision tree and ensemble classifier. In this project, we have used a comparative approach along with the new combination of features to detect voice pathologies which are laryngitis, cyst, non-fluency syndrome, and dysphonia from the SVD dataset. The combination of specific 13 MFCC (mel-frequency cepstral coefficients) features along with pitch, zero crossing rate (ZCR), spectral flux, spectral entropy, spectral centroid, spectral roll-off, and short term energy for more accurate detection of voice pathologies. It is proven that the combination of features extracted gives the best product on the audio, which split into 10 ms. Four machine learning classifiers, SVM, Naïve Bayes, decision tree and ensemble classifier for the inter classifier comparison, give 93.18, 99.45,100 and 51%, respectively. Out of these accuracies, both Naïve Bayes and the decision tree show the most promising results with a higher detection rate. Naïve Bayes and decision tree gives the highest reported outcomes on the selected set of features in the proposed methodology. The SVM has also been concluded to be the commonly used voice condition identification algorithm.

Short-term feeding behaviour sound classification method for sheep using LSTM networks

A deep learning approach using long-short term memory (LSTM) networks was implemented in this study to classify the sound of short-term feeding behaviour of sheep, including biting, chewing, bolus regurgitation, and rumination chewing. The original acoustic signal was split into sound episodes using an endpoint detection method, where the thresholds of short-term energy and average zero-crossing rate were utilized. A discrete wavelet transform (DWT), Mel-frequency cepstral, and principal-component analysis (PCA) were integrated to extract the dimensionally reduced DWT based Mel-frequency cepstral coefficients (denoted by PW_MFCC) for each sound episode. Then, LSTM networks were employed to train classifiers for sound episode category classification. The performances of the LSTM classifiers with original Mel-frequency cepstral coefficients (MFCC), DWT based MFCC (denoted by W_MFCC), and PW_MFCC as the input feature coefficients were compared. Comparison results demonstrated that the introduction of DWT improved the classifier performance effectively, and PCA reduced the computational overhead without degrading classifier performance. The overall accuracy and comprehensive F1-score of the PW_MFCC based LSTM classifier were 94.97% and 97.41%, respectively. The classifier established in this study provided a foundation for an automatic identification system for sick sheep with abnormal feeding and rumination behaviour pattern. Keywords: sheep behaviour, short-term feeding behaviour, acoustic analysis, Mel-frequency cepstral coefficients, long-short term memory networks DOI: 10.25165/j.ijabe.20211402.6081 Citation: Duan G H, Zhang S F, Lu M Z, Okinda C, Shen M X, Norton T. Short-term feeding behaviour sound classification method for sheep using LSTM networks. Int J Agric & Biol Eng, 2021; 14(2): 43–54.

Zero-crossing rate method as an efficient tool for combustion instability diagnosis

A novel methodology based on the zero-crossing rate (ZCR) is introduced to diagnose combustion instability (CI) in a combustion system. This method can assess the combustion status by counting the number of points of dynamic pressure (DP) signals crossing zero in a cycle, and quantifies the cyclic variation of the number of crossing points. To validate the feasibility of this method, transition experiments were performed in a lab-scale gas turbine combustor. The performance of the ZCR method was compared with that of a conventional method based on the root mean square (RMS) and variance (VR) for the CI detection time and stability margin. The ZCR method detected the CI 0.84 s earlier, which is an 11.11% improvement compared to the RMS method. Additionally, the ZCR method provides information regarding the stability margin between the current combustion status and CI limit, which is a crucial parameter to ensure safe operation. Furthermore, in the ZCR method, the CI frequency can be derived without fast Fourier transform calculation by analyzing the periodic characteristics of DP. Moreover, this method can assess multi-mode and single-mode CI, which makes the ZCR a potential indicator for CI monitoring systems in gas turbine combustors.

An improved intrinsic time-scale decomposition method based on adaptive noise and its application in bearing fault feature extraction

This study proposes a new method for detecting and diagnosing fault information for rolling bearings. First, a novel method called complete ensemble intrinsic timescale decomposition with adaptive noise is proposed. Compared with the complete ensemble empirical modes decomposition with adaptive noise (CEEMDAN) method, the proposed method reduces the spurious mode, which interferes with the identification of feature frequency in rolling bearings. In addition, when a typical mode mixing signal is decomposed, the mode mixing problem that appears in the adjacent mode components of the CEEMDAN method is also solved. Second, there is only an intrinsic mode function, which mostly reflects the vibration feature of a rolling bearing. We propose a new method to evaluate these mode functions to determine which mode function contains the most useful information. Our proposed method solves the problem using traditional time-domain analysis methods of a single dimensionless parameter with poor stability and low sensitivity. Our method includes sample entropy, kurtosis, the zero-crossing rate, and a correlation coefficient. The proposed method reduces the limitations of each method to select the sensitive component. Finally, we use a simulation signal and an experimental signal to verify the effectiveness of the proposed method.