Convolutional Neural Network-based Acoustic Scene Classification using Litis Rouen Dataset

MFF-SAug: Multi feature fusion with spectrogram augmentation of speech emotion recognition using convolution neural network

In Acoustic Scene Classification (ASC) two major approaches have been followed. While one utilizes engineered features such as mel-frequency-cepstral-coefficients (MFCCs), the other uses learned features that are the outcome of an optimization algorithm. I-vectors are the result of a modeling technique that usually takes engineered features as input. It has been shown that standard MFCCs extracted from monaural audio signals lead to i-vectors that exhibit poor performance, especially on indoor acoustic scenes. At the same time, Convolutional Neural Networks (CNNs) are well known for their ability to learn features by optimizing their filters. They have been applied on ASC and have shown promising results. In this paper, we first propose a novel multi-channel i-vector extraction and scoring scheme for ASC, improving their performance on indoor and outdoor scenes. Second, we propose a CNN architecture that achieves promising ASC results. Further, we show that i-vectors and CNNs capture complementary information from acoustic scenes. Finally, we propose a hybrid system for ASC using multi-channel i-vectors and CNNs by utilizing a score fusion technique. Using our method, we participated in the ASC task of the DCASE-2016 challenge. Our hybrid approach achieved 1st rank among 49 submissions, substantially improving the previous state of the art.

The aim of the study was to inspect the acoustic properties and sleep characteristics of a preapneic snoring sound. The feasibility of forecasting upcoming respiratory events by snoring sound was also investigated. Participants with habitual snoring or a heavy breathing sound during sleep were recruited consecutively. Polysomnography was conducted, and snoring-related breathing sound was recorded simultaneously. Acoustic features and sleep features were extracted from 30-second samples, and a machine learning algorithm was used to establish 2 prediction models. A total of 74 eligible participants were included. Model 1, tested by 5-fold cross-validation, achieved an accuracy of 0.92 and an area under the curve of 0.94 for respiratory event prediction. Model 2, with acoustic features and sleep information tested by Leave-One-Out cross-validation, had an accuracy of 0.78 and an area under the curve of 0.80. Sleep position was found to be the most important among all sleep features contributing to the performance of the 2 models. Preapneic sound presented unique acoustic characteristics, and snoring-related breathing sound could be deployed as a real-time apneic event predictor. The models, combined with sleep information, serve as a promising tool for an early warning system to forecast apneic events. Wang B, Yi X, Gao J, etal. Real-time prediction of upcoming respiratory events via machine learning using snoring sound signal. J Clin Sleep Med. 2021;17(9):1777-1784.

Aim: This study sought to evaluate the effects of different irrigation solutions and sterilization processes on the surface properties of ProTaper Next (PTN; Dentsply-Maillefer, Ballaigues, Switzerland) and TruNatomy (TN; Dentsply Sirona, Ballaigues, Switzerland) nickel-titanium rotary files. Methodology: 30 ProTaper Next and 30 TruNatomy files were used. Both types of files were divided into five groups. The files in three groups were dipped into 5.25% NaOCI (CanalPro, Coltene/Whaledent, Altstätten, Switzerland), BioAKT (New Tech Solutions s.r.l., Brescia, Italy), and 18% etidronic acid (Dual Rinse HEDP, Medcem, Weinfelden, Switzerland), respectively, for 5 minutes, while the files in the fourth group were exposed to five autoclave cycles. The files in the fifth (control) group did not receive any treatment. Scans were performed from five different points using an atomic force microscope (AFM). The assumption of the conformity of the quantitative variables to a normal distribution was evaluated using the Kolmogorov-Smirnov test. A one-way analysis of variance was used to compare the quantitative measures between more than two independent groups, while the post-hoc least significant difference test was used for the pairwise comparisons (p˂0.05). Results: AFM images of the PTN and TN files in all the groups revealed surface irregularities. The root mean square (RMS) and depth values of the files dipped in NaOCI, BioAKT, and Etidronic acid were statistically higher than those of the control group (p<0.05). Moreover, the RMS and depth values were statistically higher in the files that had been subjected to five autoclave cycles when compared with the control group (p<0.05). The RMS values and depth values among the groups were found to be compatible. Conclusion: All the applications increased the surface roughness of the PTN and TN files. The Etidronic acid and sterilization applications were found to increase the RMS and depth values ​​the most when compared with the control group. How to cite this article: Öztekin F, Elmalı E. The effect of irrigation solutions and sterilization process on the surface properties of ProTaper Next and TruNatomy rotary files: An atomic force microscopy study. Int Dent Res 2022;12(2):39-46. https://doi.org/10.5577/intdentres.2022.vol12.no2.1 Linguistic Revision: The English in this manuscript has been checked by at least two professional editors, both native speakers of English.

An ensemble 1D-CNN-LSTM-GRU model with data augmentation for speech emotion recognition

In the project, an emotion recognition system from speech is proposed using deep learning. The goal of this project is to classify a speech signal into one of the five emotions listed below: anger, boredom, fear, happiness, and sadness. Snippets below from numerous Marathi movies and TV shows were used to construct the dataset for Marathi language samples which include 20 audio samples for anger, 19 for boredom, 5 for fear, and 11 for happiness. The proposed system first processes a speech signal from the time domain to the frequency domain using Discrete Time Fourier Transform (DTFT). Then, data augmentation is performed which includes noise injection, stretching, shifting, and pitch scaling of the speech signal. Next, feature extraction is performed in which 5 features were selected, which include Mel Frequency Cepstral Coefficients (MFCC), Zero Crossing Rate (ZCR), Chroma STFT, Mel Spectrogram, and Root mean square value. These features were then fed to a Convolutional Neural Network (CNN). The efficiency of the suggested system employing the CNNs is supported by experimental findings. This model’s accuracy on the test data is 80.33%, and its f1 values for anger, boredom, fear, happiness, and sadness are 0.85, 0.83, 0.50, 0.62, and 0.84, respectively.

Inhomogeneities in muscle activation reveal motor unit recruitment

Acoustic scene classification is the task of identifying the scene from which the audio signal is recorded. Convolutional neural network (CNN) models are widely adopted with proven successes in acoustic scene classification. However, there is little insight on how an audio scene is perceived in CNN, as what have been demonstrated in image recognition research. In the present study, the Class Activation Mapping (CAM) is utilized to analyze how the log-magnitude Mel-scale filter-bank (log-Mel) features of different acoustic scenes are learned in a CNN classifier. It is noted that distinct high-energy time-frequency components of audio signals generally do not correspond to strong activation on CAM, while the background sound texture are well learned in CNN. In order to make the sound texture more salient, we propose to apply the Difference of Gaussian (DoG) and Sobel operator to process the log-Mel features and enhance edge information of the time-frequency image. Experimental results on the DCASE 2017 ASC challenge show that using edge enhanced log-Mel images as input feature of CNN significantly improves the performance of audio scene classification.

Several exercise methods, such as the Shaker exercise, tongue press exercise, chin tuck against resistance (CTAR) exercise, and submandibular push exercise, have been introduced to strengthen the muscles involved in swallowing. In this study, we compared the effectiveness of the CTAR, submandibular push, and Shaker exercises for the induction of selective supra- and infrahyoid muscle contractions using surface electromyography (EMG). This study is a prospective non-randomized controlled study. Twenty-five healthy subjects and 20 patients experiencing swallowing difficulty were enrolled. During the three different types of exercises, the root mean square (RMS) values of the sternocleidomastoid (SCM), suprahyoid (anterior belly of the digastric and mylohyoid muscles), and infrahyoid (sternothyroid and thyrohyoid muscles) muscles were analyzed using surface EMG. Differences in the activity of swallowing muscles among the three different exercises were analyzed using one-way repeated measured analysis of variance. In terms of both the maximum and mean RMS values of the suprahyoid muscle, the submandibular push exercise showed a larger RMS value than the CTAR and Shaker exercises in healthy subjects (p < 0.05). In terms of both the maximum and mean RMS values of the suprahyoid muscle, the Shaker exercise and submandibular push exercise showed a larger RMS value than the CTAR exercise in patients with swallowing difficulty (p < 0.05). The submandibular push exercise may be effective as a swallowing muscle exercise owing to its superiority in inducing selective contractions of the supra- and infrahyoid muscles. The CTAR and Shaker exercises are also effective in this regard.

In recent years, real-time global ionospheric map (RT-GIM) products have been actively developed by the international global navigation satellite system (GNSS) service (IGS) and its ionosphere associate analysis centers (IAACs) along with the increase of RT-GNSS multi-frequency and multi-constellation observations. In this study, the accuracy and consistency of three RT-GIM products from the Chinese Academy of Sciences (CAS), Wuhan University (WHU), and IGS are evaluated and analyzed utilizing three validation methods, namely, comparison with JASON-3 vertical total electron content (VTEC), the difference of slant total electron content (dSTEC), and IGS combined final GIM (IGSG) data. The test period was from 1 January 2019 to 31 December 2022, including the different solar activities. First, the comparison with JASON-3 data illustrates that the quality of the three RT-GIM products over oceans is in great consistency with that of the IGSG during different levels of solar activity and the daily mean bias (MEAN) values in low and high solar activities are approximately 5 and 10 TECU, respectively. The root mean square (RMS) values under low and high solar activities can be up to 7 and 12 TECU. Furthermore, the dSTEC validation results present that the MEAN values of RT-GIM products from different IAACs at high- and mid-latitude stations are about 0.5 TECU, which is smaller than those at low-latitude stations at about 1 TECU over continental regions. The standard deviation (STD) and RMS values for various RT-GIM products are within 3 and 4 TECU at low latitudes, respectively. In terms of the comparison with IGSG, the result shows that IGS combined RT-GIM (IRTG) presents better consistency than CAS RT-GIM (CRTG) and WHU RT-GIM (WRTG) in 2021 and 2022, with average annual STD and RMS values of 2.56 and 2.78 TECU, respectively. The daily biases of the RT-GIM products relative to IGSG can reach 4 TECU in high solar activities and the daily STD and RMS values are mainly within the 5 to 6 TECU range, respectively.

ObjectivesTo evaluate the effect of additive and subtractive manufacturing on the accuracy (trueness and precision) of fixed partial denture patterns (FPDPs) used for casting or pressing. Materials and MethodsA 3-unit complete coverage FPD on mandibular right first premolar and first molar teeth was virtually designed. Using the design data, FPD patterns were fabricated from an additively manufactured resin (PR, ProArt Print Wax) and 2 CAD-CAM wax discs (YW, ProArt CAD Wax Yellow and BW, ProArt CAD Wax Blue) (n = 10). Each pattern was then digitized with a scanner (CEREC Primescan) and evaluated for 3D surface deviation at 4 different surfaces (overall, external, marginal, and intaglio surfaces) by using a 3D analysis software (Medit Link). Root mean square (RMS) values were automatically calculated. Data were analyzed by using Kruskal-Wallis and Dunn's post hoc tests for trueness and precision (α= 0.05). ResultsSignificant differences were found among the RMS values for overall (P<.001) and each surface (P≤.040) evaluated. PR had the highest overall (P≤.011) and intaglio surface (P≤.01) deviations, while the difference between YW and BW was not significant (P≥.199). PR had the highest (P≤.027) and BW had the lowest (P≤.042) external surface mean RMS values. BW had higher mean marginal RMS value than YW (P=.047). For precision, significant differences were observed among test groups only for marginal RMS values (P=.002). PR had lower precision than BW (P=.002). ConclusionsBW and YW FPDPs mostly had higher trueness compared with PR FPDPs. However, considering relatively smaller deviations at marginal and intaglio surfaces and the fact that patterns mostly had similar precision, clinical fit of FPDs fabricated by using tested patterns may be similar. Clinical SignificanceDefinitive 3-unit fixed partial dentures fabricated by using tested patterns may be similar. However, FPDs fabricated with tested additively manufactured resin patterns might result in more chairside adjustments than those fabricated with tested subtractively manufactured wax patterns.

The objective of the present study is to develop a vibration fatigue life prediction model for ball grid array (BGA) solder joint. In this study, 3D global/local finite element models were first constructed with MSC/PATRAN code. For the global model including BGA package soldered onto the printed wiring board (PWB), linear finite element dynamic analyses with an excitation normal to the PWB were conducted using MSC/NASTRAN code to determine the dynamic responses at the two ends of BGA solder joint. In these analyses, a single-degree-of-freedom system was assumed. A steady-state harmonic (or frequency sweep) response provides phase angle values while a random response furnishes the power spectral densities as well as their root mean square (RMS) values. For the local model having a refined mesh to simulate the local region in detail, a linear finite element static analysis was conducted by applying the derived RMS values with their corresponding phase angles to this model (at the two ends of the critical corner solder joint) in order to determine the solder effective stress/strain RMS value. A vibration fatigue life model, evolved from an empirically derived formula of universal slopes based on high-cycle fatigue test data, was established. This model combined with a three-band technique and the derived solder effective strain was then used to predict the BGA solder joint survivability/durability. This prediction will be compared to test results, which will be provided by in-house on-going R&D, to validate the proposed BGA solder joint vibration fatigue life prediction model. An example of a 600-pin plastic BGA soldered onto the various locations of the PWB was illustrated in the present study.

The statistical properties of acoustic emission signals for tool condition monitoring (TCM) applications in mechanical lathe machining are analyzed in this paper. Time series data and root mean square (RMS) values at various tool wear levels are shown to exhibit features that can be put into relation with aging in both cases. In particular, the histograms of raw data show power-law distributions above a cross-over value, in which newer cutting tools exhibit more numerous larger events compared with more worn-out ones. For practical purposes, statistics based on RMS values are more feasible, and the analysis of these also reveals discriminating age-related features. The assumption that experimental RMS histograms follow a Beta (β) distribution has also been tested. The residuals of the modeling β functions indicate that the search for a more appropriate fitting function for the experimental distribution is desirable.

The aim of this study was to evaluate the effect of glide path preparation on the surface topography changes of WaveOne and OneShape single files using atomic force microscopy analysis. Two commercially available brands of single file systems were analyzed, including 15 specimens/brand as following: Group I; WaveOne 0.08/25 (Dentsply Maillefer, Ballaigues, Switzerland) and Group II; OneShape 0.06/25 (Micro Mega, Besançon, France). In each group, 3 new files were subjected to AFM analysis as received from manufacturer, and 12 files were analyzed after preparation of root canals. A total of 20 human maxillary molar teeth were selected. In each group, 12 files were divided into 4 subgroups according to instruments that were used in one or three root canals with or without a previous glide path (n = 3). After preparation the surface analysis of single files was performed on 11 different points located along a 3-mm section starting at the tip of each file using AFM. Root mean square (RMS) values were recorded to investigate the surface features of endodontic files. The differences between RMS values were tested by one-way ANOVA. There was no statistically significant difference between the groups (p > 0.05). Creation of a glide path did not affect the surface topography of single file systems.

Wavefront aberrations (WFAs) of the eye vary with time because of the tear film dynamics. We investigated, using a simulation method, the variation of optical quality with time-varying wavefront measurements of 13 eyes with different refractions. WFAs of 13 normal eyes of 13 subjects were measured every second for 10 s. First, we simulated WFAs with conventional corneal laser refractive surgery by subtracting the second-order aberrations of the least aberrated measurement from measured consecutive WFAs. Second, we simulated customized refractive surgery by subtracting the second- to sixth-order aberrations of the least aberrated measurement from measured consecutive WFAs. We calculated Strehl ratios and retinal images from these corrected consecutive WFAs. In one eye, the root mean square (RMS) values of WFAs with a second-order correction were sometimes smaller than those of WFAs with a second- to sixth-order correction, when these were compared at the same time point after a blink. However, in the other 12 eyes, the RMS values with second- to sixth-order corrections were smaller than those with only a second-order correction. In eight eyes, the Strehl ratios with second- to sixth-order corrections were larger than those with second-order corrections. In the remaining five eyes, Strehl ratios with second- to sixth-order corrections were sometimes smaller than those with second-order corrections. In a simulation, the correction of time-invariant higher order aberrations usually reduced RMS values, but it did not always result in higher Strehl ratios than those obtained with only second-order corrections.