dB SNR Research Articles

Speech Emotion Recognition under Noisy Environments with SNR Down to −6 dB Using Multi-Decoder Wave-U-Net

A speech emotion recognition (SER) model for noisy environments is proposed, by using four band-pass filtered speech waveforms as the model input instead of the simplified input features such as MFCC (Mel Frequency Cepstral Coefficients). The four waveforms retain the entire information of the original noisy speech while the simplified features keep only partial information of the noisy speech. The information reduction at the model input may cause the accuracy degradation under noisy environments. A normalized loss function is used for training to maintain the high-frequency details of the original noisy speech waveform. A multi-decoder Wave-U-Net model is used to perform the denoising operation and the Wave-U-Net output waveform is applied to an emotion classifier in this work. By this, the number of parameters is reduced to 2.8 M for inference from 4.2 M used for training. The Wave-U-Net model consists of an encoder, a 2-layer LSTM, six decoders, and skip-nets; out of the six decoders, four decoders are used for denoising four band-pass filtered waveforms, one decoder is used for denoising the pitch-related waveform, and one decoder is used to generate the emotion classifier input waveform. This work gives much less accuracy degradation than other SER works under noisy environments; compared to accuracy for the clean speech waveform, the accuracy degradation is 3.8% at 0 dB SNR in this work while it is larger than 15% in the other SER works. The accuracy degradation of this work at SNRs of 0 dB, −3 dB, and −6 dB is 3.8%, 5.2%, and 7.2%, respectively.

Automatic classification of multi-carrier modulation signal using STFT spectrogram and deep CNN

In the realm of communication systems, categorizing Multi-Carrier Modulation (MCM) signals without cooperative communication poses a significant technical challenge. In this paper, we introduce a novel approach for accurately categorizing five distinct MCM signals, including Orthogonal Frequency Division Multiplexing (OFDM), Filter Bank Multicarrier (FBMC), Filtered Orthogonal Frequency Division Multiplexing (FOFDM), Windowed Orthogonal Frequency Division Multiplexing (WOLA), and Universal Filtered Multicarrier (UFMC). Each signal is considered with two types of subcarrier waveforms, Quadrature Amplitude Modulation 16 (QAM16) and Quadrature Amplitude Modulation 64 (QAM64), resulting in a total of 10 unique MCM signals for classification. Our proposed methodology leverages Short-Time Fourier Transform (STFT) spectrograms for feature extraction at the frontend, while at the backend, we employ three variants of Convolutional Neural Network (CNN) models; CNN, CNN with Dropout (CNN_d), CNN with both Dropout and L1 Regularization (CNN_dL1) and one deep CNN model; Xception, individually. We aim to provide an efficient and reliable means of categorizing MCM signals, with practical applications in signal processing and communication systems. Extensive simulations demonstrate the effectiveness of our approach, achieving remarkable accuracies. Notably, the Xception model exhibits the highest accuracy among the four models considered. Specifically, we attain an accuracy of 98% at 10 dB SNR using the Xception model. These results underscore the efficacy of our proposed methodology and highlight the potential for its deployment in real-world scenarios.

Deep learning based detection of signal of interest of uplink NOMA multi-users in NOMA In-band Full-duplex Radio communication

The research focuses on investigating the symbol error rate (SER) performance of the Non Orthogonal Multiple Access (NOMA) In-band Full-duplex Radio communication system, which is considered a promising technology for 5G/6G. This system facilitates simultaneous transmission and reception of data over the same resource block for multiple users via power domain multiplexing. However, the system faces a significant challenge in the form of self-interference signal (SI). The SI signal transmitted from the NOMA In-band full-duplex radio transmitting antenna is also received by its own receiving antenna. It limits perfect successive interference cancellation (SIC) for channel estimation and signal detection of the signal of interest (SoI) from uplink NOMA multi-users at the receiver of NOMA In-band Full-duplex radio. Furthermore the SER performance is significantly degraded due to various SI and SoI channel imperfections. To address this, a three-phase-based efficient approach has been proposed for simultaneous non-linear SI cancellation and detection of the SoI from uplink NOMA two users under aforementioned SI and SoI channel imperfections. The approach uses the Least Square-SIC Estimator, Minimum Mean Square Error-SIC Estimator, and deep learning (DL) methods. The reported 4 dB SNR gain compared to least square-SIC/minimum mean square error SIC and DNN-based methods demonstrate the robustness of the proposed DL-aided approach against various SI and SoI channel imperfections for NOMA In-band Full-duplex Radio communication.

Impact of perceptual-learning strategies and background noise on disordered speech intelligibility

Objective: There is a plethora of research showing reduced speech intelligibility for a variety of voice disorders (i.e., dysphonia, alaryngeal). Therapeutic approaches to improve intelligibility typically involve targeting the speaker (e.g., clear speech, reduced rate) with minimal attention to the listener. Therefore, there were three purposes of this study: 1) to determine the impact of background noise on the speech intelligibility of disordered speakers; 2) to determine the impact of providing listeners with perceptual-learning strategies on the speech intelligibility of speakers with a voice disorder; and 3) to determine if subjective ratings of voice quality can predict speech intelligibility. Methods: Sentences were recorded from 12 speakers (2 typical, 3 alaryngeal, 7 dysphonic). Sentences were divided into one of three groups of signal-to-noise ratios (SNR: quiet, +5 dB SNR, and 0 dB SNR) and individually presented to 129 healthy listeners divided into one of three groups (i.e., control, acknowledgment of disorder, cognitive-perceptual strategies). Orthographic transcription was used to assess speech intelligibility. In addition, three expert listeners provided subjective voice quality ratings of all speakers. Results: Listeners had significantly more intelligibility errors with increased background noise (p <.001) and providing strategies to listeners did not result in a statistical improvement level F(6, 486) = 1.53, p = .17, η2p = 0.02. Regression analysis showed that the subjective voice quality overall severity was able to predict speech intelligibility in the noisy condition (0 dB SNR), accounting for 37% of the variance, R2 = .365, F(1,10) = 5.759, p = .037. Conclusion: Results suggest that increased background noise has a deleterious effect on the speech intelligibility of those with a voice disorder but that providing listeners with strategies in hopes of improving speaker intelligibility was not successful. Results did provide support, however, for the use of subjective voice quality ratings as a potential index of speech intelligibility.

Speech discrimination with separated signal sources and sound localization with speech stimuli : Learning effects and reproducibility

Binaural hearing enables better speech comprehension in noisy environments and is necessary for acoustic spatial orientation. This study investigates speech discrimination in noise with separated signal sources and measures sound localization. The aim was to study characteristics and reproducibility of two selected measurement techniques which seem to be suitable for description of the aforementioned aspects of binaural hearing. Speech reception thresholds (SRT) in noise and test-retest reliability were collected from 55normal-hearing adults for aspatial setup of loudspeakers with angles of ± 45° and ± 90° using the Oldenburg sentence test. The investigations of sound localization were conducted in asemicircle and fullcircle setup (7 and 12 equidistant loudspeakers). SRT (S-45N45: -14.1 dB SNR; S45N-45: -16.4 dB SNR; S0N90: -13.1 dB SNR; S0N-90: -13.4 dB SNR) and test-retest reliability (4 to 6 dB SNR) were collected for speech intelligibility in noise with separated signals. The procedural learning effect for this setup could only be mitigated with 120 training sentences. Significantly smaller SRT values, resulting in better speech discrimination, were found for the test situation of the right compared to the left ear. RMS values could be gathered for sound localization in the semicircle (1,9°) as well as in the fullcircle setup (11,1°). Better results were obtained in the retest of the fullcircle setup. When using the Oldenburg sentence test in noise with spatially separated signals, it is mandatory to perform atraining session of 120sentences in order to minimize the procedural learning effect. Ear-specific SRT values for speech discrimination in noise with separated signal sources are required, which is probably due to the right-ear advantage. Atraining is recommended for sound localization in the fullcircle setup.

Analysis of Filtered Multicarrier Modulation Techniques Using Different Windows for 5G and Beyond Wireless Systems

In contemporary wireless communication systems, multicarrier modulation schemes have become widely adopted over single-carrier techniques due to their improved capacity to address challenges posed by multipath fading channels, leading to enhanced spectral efficiency. Orthogonal frequency division multiplexing (OFDM), a prevalent multicarrier scheme in 4G, is favored for its ease of implementation, interference resilience, and high data rate provision. But it falls short of meeting the requirements for 5G and beyond due to limitations such as out-of-band (OOB) emissions and cyclic prefixes. This paper introduces the filter bank multicarrier (FBMC) and universal filtered multicarrier (UFMC) with quadrature amplitude modulation (QAM) and phase shift keying (PSK) waveforms through Additive White Gaussian Noise channel (AWGN), Rayleigh fading channel and Rician channel. The objective of this paper is to enhance the performance of UFMC with reduced complexity through the new filtering approach for achieving optimal outcomes. The proposed scheme, incorporating Tukey filtering technique, demonstrates superior performance in reducing peak-to-average power ratio (PAPR) and improving bit error ratio (BER) compared to the original UFMC signal without necessitating additional power increments. Specifically, the UFMC system with Tukey filtering achieves a notable net gain of 5 dB. Simulation results demonstrate that utilizing various filter types in FBMC and UFMC systems, combined with QAM modulation, significantly reduces OOB emissions compared to conventional systems. In aspect to BER, Tukey window showed almost 10−6 at 15 dB SNR in UFMC which is better than FBMC.

Intelligibility of British and American English in different listening conditions

Communication specialists frequently hear complaints from patients that unfamiliar accents are more difficult to understand, particularly in background noise. Previous work (Schilaty et al., 2021) has indeed shown that when presenting American listeners with both American and British accents, intelligibility in noise is poorer for the unfamiliar accent across age and hearing status. However, Schilaty et al. (2021) examined relatively easy sentence materials (Basic English Lexicon [BEL] sentences; Calandruccio & Smiljanic, 2012) in only one type of noise. The current study expands on previous work by comparing the intelligibility of American and British accents for a more difficult sentence set (Harvard sentences; Rothauser et al., 1969) in young adult American listeners across two noise types (speech-shaped noise and 12-talker babble). Stimuli consisted of 180 Harvard sentences produced by two male talkers: one with an American (Northwestern) dialect and one with a Southern British dialect. Young adult listeners with normal hearing transcribed the sentences in quiet, in speech-shaped noise at −3 dB SNR, and in 12-talker babble at −3 dB SNR. Listeners also completed a dialect familiarity questionnaire. The results will shed additional light on the interactions between talker and listener dialect, dialect familiarity, and background noise type.

Effect of subsequent context on the real-time interpretation of ambiguous target words in spectrally degraded speech

Subsequent context in a sentence can be used to revise the interpretation of a word that was not clearly heard, a frequent occurrence in degraded speech. One strategy for avoiding the cognitive cost associated with revising an initial interpretation is to delay commitment to an interpretation until the end of the sentence, sometimes called a “wait-and-see” strategy. Listening with a significant hearing loss (i.e., constantly hearing degraded speech), may prompt a listener to use this strategy even in relatively easy listening conditions. This study used an eye-tracker to record gaze patterns during a phoneme classification task of ambiguous target words in two levels of background noise ( + 5 and −5 dB SNR). Participants either used cochlear implants (CIs) or were presented 8-channel noise vocoded speech. The target words were at the beginning of sentences; the subsequent context cued one target word interpretation (based on phoneme classification) or the other. We hypothesized that listeners who use CIs will demonstrate a delaying strategy at both levels of background noise, while listeners who have acoustic hearing will only demonstrate a delaying strategy at the highest level of background noise. These findings have implications for rehabilitative strategies to improve communication outcomes in adults with CIs.

Predicting Early Cochlear Implant Performance: Can Cognitive Testing Help?

There is significant variability in speech recognition outcomes in adults who receive cochlear implants (CIs). Little is known regarding cognitive influences on very early CI performance, during which significant neural plasticity occurs. Prospective study of 15 postlingually deafened adult CI candidates tested preoperatively with a battery of cognitive assessments. The mini-mental state exam (MMSE), forward digit span, Stroop measure of inhibition-concentration, and test of word reading efficiency were utilized to assess cognition. consonant-nucleus-consonant words, AZBio sentences in quiet, and AZBio sentences in noise (+10 dB SNR) were utilized to assess speech recognition at 1- and 3-months of CI use. Performance in all speech measures at 1-month was moderately correlated with preoperative MMSE, but these correlations were not strongly correlated after correcting for multiple comparisons. There were large correlations of forward digit span with 1-month AzBio quiet (P ≤ 0.001, rho = 0.762) and AzBio noise (P ≤ 0.001, rho = 0.860), both of which were strong after correction. At 3 months, forward digit span was strongly predictive of AzBio noise (P ≤ 0.001, rho = 0.786), which was strongly correlated after correction. Changes in speech recognition scores were not correlated with preoperative cognitive test scores. Working memory capacity significantly predicted early CI sentence recognition performance in our small cohort, while other cognitive functions assessed did not. These results differ from prior studies predicting longer-term outcomes. Findings and further studies may lead to better preoperative counseling and help identify patients who require closer evaluation to ensure optimal CI performance.

Coupling quantitative ultrasound using echo envelop statistics and shear wave propagation to provide new image contrast of mimicked liver lesions

In quantitative ultrasound, biomarkers are sensitive to the scatterers’ positioning and number density, and therefore, shear wave (SW) propagation may modify their time-varying behaviors and provide new image contrast. Herein, we considered echo envelop statistics by using one parameter of the homodyned-K distribution (HKD): the diffuse-to-total signal power ratio 1/(κ+1). A liver lesion-mimicking phantom was simulated with an inclusion and a background of elastic moduli of 40 and 4 kPa, and scatterers’ number densities per resolution cell of 15 and 5, respectively. Ten sets of spatially correlated scatterers were created, and a plane SW was simulated to reposition time and spatially varying scatterers according to the SW motion. Ultrasound images were simulated using MUST with added Gaussian noise to attain 20-dB SNR. Dynamic and static analyses were made over 30 frames with and without SW motion, respectively. Validation was based on the contrast of 1/(κ+1) between the inclusion and the background. HKD image contrast increased with SW motion with values from 0.12 ± 0.10 to 0.48 ± 0.16 (no units) (p < 0.001). Results suggest that the variation in scatterers’ organization under SW propagation may provide new information over its static counterpart to enhance the contrast of liver lesions.

Effects of time compressed speech and background noise on memory recall

Listeners demonstrate remarkable flexibility in processing speech under varying conditions, rapidly adapting to acoustically distorted signals like time-compressed speech or speech in noise. Existing research has explored the immediate effects of time-compression and background noise on speech intelligibility, but the impact of these distortions on long-term memory recall for speech content is less well understood. With the growing use of online education platforms where listeners often prefer to use accelerated playback, understanding the effects of time-compression and noise on long-term memory from speech is crucial. Here, we investigated how speech distortions from time-compression influence explicit memory, and how this interacts with background noise. Participants (N = 35) each listened to six recorded lectures at varying time-compression rates (1.0x, 1.5x, 2.0x) and background noise levels (quiet versus babble noise at + 10 dB SNR). After each lecture, participants answered 10 multiple-choice questions probing explicit content recall. Listeners exhibited a significant decline in content recall with increased time compression, while background noise during the lecture did not impact listeners’ content recall. Our findings demonstrate that time-compressed speech negatively affects long-term memory recall, which may arise from the additional cognitive demands necessary to process time-compressed speech, even after perceptual adaptation.

Listening effort mitigates rollover effects on speech-in-noise perception

Increasing the sound intensity may lead to worse speech understanding, especially in noise. This is known as the “Rollover” phenomenon. There is mounting evidence that listening effort plays an important role in challenging listening conditions and can be directly quantified with objective measures such as pupil dilation. However, there is limited understanding of how listening effort relates to rollover in speech understanding. We hypothesized that listening effort plays an essential role in mitigating rollover effects to differential extents across age and hearing status. We recruited across the adult lifespan (N = 50, 20–83 years) with different hearing statuses in acoustic listeners and cochlear implant users to perform a speech discrimination task. Minimal word pairs were presented both in quiet and in 0 dB SNR babble noise, ranging from 35–85 dB SPL. Pupil area was tracked simultaneously with behavioral responses during the task. We found that normal-hearing listeners are fully able to utilize effort contributions to minimize rollover effects between in quiet and in noise conditions, with diminishing benefit as a function of age and increased hearing loss. The results of this project could broadly influence how to design future hearing devices and interventions that maximize hearing abilities for those affected by hearing loss.

Listening Effort for Speech in Noise Perception Using Pupil Dilation: A Comparison Among Percussionists, Non-Percussionists, and Non-Musicians.

Most studies in literature attribute the benefits of musical training on speech in noise (SIN) perception to "experience-based" plasticity, which assists in the activation of speech-processing networks. However, whether musicianship provides an advantage for the listening effort (LE) required to comprehend speech in degraded environments has received less attention. The current study aimed to understand the influence of Indian classical music training on SIN perception and its related LE across percussionists, non-percussionists, and non-musicians. A quasi-experiment was conducted on 16 percussionists, 17 non-percussionists, and 26 non-musicians aged 18-35 years with normal hearing. In phase 1, musical abilities were assessed using Mini-Profile of Music Perception Skills (Mini-PROMS). Phase 2 examined SIN using Tamil Phonemically-Balanced Words and Tamil Matrix Sentence Test at +5 dB, 0 dB, and -5 dB SNR and LE using pupillometry, measuring pupil dilations with an eye-tracker. Fractional Logit and Linear Regression models demonstrated that percussionists outperformed non-percussionists in Tuning and Speed subsets of Mini-PROMS. Percussionists outperformed non-percussionists and non-musicians in SIN and LE at -5 dB SNR for words and at 0 dB and -5 dB SNR for sentences. Percussionists have the greatest advantage in decoding SIN with reduced LE followed by non-percussionists and non-musicians, demonstrating musician-advantage in most challenging listening conditions.

Do Different Types of Microphones Affect Listening Effort in Cochlear Implant Recipients? A Pupillometry Study.

It is known that subjects with a cochlear implant (CI) need to exert more listening effort to achieve adequate speech recognition compared to normal hearing subjects. One tool for assessing listening effort is pupillometry. The aim of this study is to evaluate the effectiveness of adaptive directional microphones in reducing listening effort for CI recipients. We evaluated listening in noise and listening effort degree (by pupillometry) in eight bimodal subjects with three types of CI microphones and in three sound configurations. We found a correlation only between sound configurations and listening in noise score (p-value 0.0095). The evaluation of the microphone types shows worse scores in listening in noise with Opti Omni (+3.15 dB SNR) microphone than with Split Dir (+1.89 dB SNR) and Speech Omni (+1.43 dB SNR). No correlation was found between microphones and sound configurations and within the pupillometric data. Different types of microphones have different effects on the listening of CI patients. The difference in the orientation of the sound source is a factor that has an impact on the listening effort results. However, the pupillometry measurements do not significantly correlate with the different microphone types.

Frankfurt concept of early inpatient rehabilitation after cochlear implant treatment : Feasibility for aftercare. German version

The Association of the Scientific Medical Societies in Germany (AWMF) clinical practice guideline on cochlear implant (CI) treatment, which was updated in 2020, defined the entire process of CI care for the first time. In the present study, the feasibility and results of very early rehabilitation were examined. The intervention group (IG) comprised 54patients in whom rehabilitation was initiated within 14(maximally28) days after implantation. Patients with asignificantly longer waiting time were included in the control group (CG, n = 21). In addition to the start and duration of rehabilitation, the speech intelligibility achieved with CI was recorded at different timepoints within a12-month period. In addition, questionnaires were used to assess the effort of fitting the CI processor and the patients' satisfaction with the outcome as well as the timing of the start of rehabilitation. Median waiting time between implantation and start of rehabilitation was 14days in the IG and 106days in the CG; 92.6% of IG patients were able to start rehabilitation within 14days. The effect of rehabilitation in the IG was 35and in the CG 25percentage points (Freiburg monosyllabic test). After 6and 12months of CI use, both groups showed comparable results in the test condition in quiet (IG/CG6months: 70%/70%; 12months: 70%/60%, Freiburg monosyllabic test) and in noise (IG/CG6months: -1.1-0.85 dB SNR; 12months: -0.65 dB SNR/0.3 dB SNR, Oldenburg sentence test). Hearing quality assessment scores collected by SSQ (Speech, Spatial and Qualities of Hearing Scale) questionnaire showed better scores in the IG at 6months, which converged to CG scores at 12months. The IG was significantly more satisfied with the timing of the start of rehab than the CG. All other data obtained from questionnaires showed no differences between the two groups. Avery early start of inpatient rehabilitation after cochlear implantation was successfully implemented. The rehabilitation was completed within 7weeks of CI surgery. Comparison of speech recognition test results before and after rehabilitation showed asignificant improvement. Aclear rehabilitation effect can therefore be demonstrated. Inclusion of CI rehabilitation in the German catalog of follow-up treatments is thus scientifically justified and therefore strongly recommended.

Predictors for estimating subcortical EEG responses to continuous speech.

Perception of sounds and speech involves structures in the auditory brainstem that rapidly process ongoing auditory stimuli. The role of these structures in speech processing can be investigated by measuring their electrical activity using scalp-mounted electrodes. However, typical analysis methods involve averaging neural responses to many short repetitive stimuli that bear little relevance to daily listening environments. Recently, subcortical responses to more ecologically relevant continuous speech were detected using linear encoding models. These methods estimate the temporal response function (TRF), which is a regression model that minimises the error between the measured neural signal and a predictor derived from the stimulus. Using predictors that model the highly non-linear peripheral auditory system may improve linear TRF estimation accuracy and peak detection. Here, we compare predictors from both simple and complex peripheral auditory models for estimating brainstem TRFs on electroencephalography (EEG) data from 24 participants listening to continuous speech. We also investigate the data length required for estimating subcortical TRFs, and find that around 12 minutes of data is sufficient for clear wave V peaks (>3 dB SNR) to be seen in nearly all participants. Interestingly, predictors derived from simple filterbank-based models of the peripheral auditory system yield TRF wave V peak SNRs that are not significantly different from those estimated using a complex model of the auditory nerve, provided that the nonlinear effects of adaptation in the auditory system are appropriately modelled. Crucially, computing predictors from these simpler models is more than 50 times faster compared to the complex model. This work paves the way for efficient modelling and detection of subcortical processing of continuous speech, which may lead to improved diagnosis metrics for hearing impairment and assistive hearing technology.

The influence of the stimulus level used to prescribe nonlinear frequency compression on speech perception.

Nonlinear frequency compression (NFC) is a signal processing technique designed to lower high frequency inaudible sounds for a listener to a lower frequency that is audible. Because the maximum frequency that is audible to a listener with hearing loss will vary with the input speech level, the input level used to set nonlinear frequency compression could impact speech recognition. The purpose of this study was to determine the influence of the input level used to set nonlinear frequency compression on nonsense syllable recognition. Nonsense syllable recognition was measured for three NFC fitting condition (i.e., with nonlinear frequency compression set based on speech input levels of 50-, 60-, and 70-dB SPL, respectively), as well as without nonlinear frequency compression (restricted bandwidth condition). Twenty-three adults (ages 42-80 years old) with hearing loss. Data were collected, monaurally, using a hearing aid simulator. The start frequency and frequency compression ratios were set based on the SoundRecover Fitting Assistant. Speech stimuli were 657 consonant-vowel-consonant nonwords presented at 50, 60, and 70 dB SPL and mixed with steady noise (6 dB SNR) and scored based on entire word, initial consonant, vowel, and final consonant. Linear mixed effects examined the effects of NFC fitting condition , presentation level, and scoring method on percent correct recognition. Additional predictor variables of start frequency and frequency-compression ratio were examined. Nonsense syllable recognition increased as presentation level increased. Nonsense syllable recognition for all presentation levels was highest when nonlinear frequency compression was set based on the 70 dB SPL input level and decreased significantly when set based on the 60- and 50-dB SPL inputs. Relative to consonant recognition, there was a greater reduction in vowel recognition. Nonsense syllable recognition between NFC fitting conditions improved with increases in the start frequency, where higher start frequencies led to better nonsense word recognition. Nonsense syllable recognition was highest when setting nonlinear frequency compression based on a 70 dB SPL presentation level and suggest that a high presentation level should be used to determine nonlinear frequency compression parameters for an individual patient.

Development and validation of a French speech-in-noise self-test using synthetic voice in an adult population

BackgroundSpeech-in-noise (SIN) audiometry is a valuable part of audiological diagnostics and clinical measurements, providing information on an individual's ability to understand in noise. To date, such tests have been developed with natural voice presented monaurally or binaurally (via diotic and dichotic presentation). The time required to develop such tests is often long. The use of synthetic voice would simplify the test creation process and self-testing would improve accessibility.DesignMeasurements were performed using an Android tablet (Samsung Galaxy Tab A7) and calibrated Bluetooth headphones (Tilde Pro C, Orosound). Normative values were first defined using natural or synthetic voice on 69 normal-hearing participants. A total of 463 participants then undertook the SIN test comprising synthetic voice and dichotic antiphasic presentation. Of these, 399 also performed the SIN test with diotic presentation.ResultsNo significant differences in the speech reception threshold (SRT) were found between natural and synthetic voices for diotic presentations (p = 0.824, paired Student t-test) with normative values of −10.7 dB SNR (SD = 1.5 dB) and −10.4 dB SNR (SD = 1.4 dB), respectively. For the SoNoise test with synthetic voice and dichotic antiphasic presentation, the results showed a normative value of −17.5 dB SNR (SD = 1.5 dB), and a strong correlation (r = 0.797, p < 0.001) with the four-frequency pure-tone average (4f-PTA). Receiver operating curves (ROC) were then calculated: for a 4f-PTA of 20 dB hearing level (HL), the SRT was −14.5 dB SNR with a sensitivity of 84% and specificity of 89%. For a 4f-PTA of 30 dB HL, the SRT was −13.7 dB SNR with a sensitivity of 89% and specificity of 91%. For a 4f-PTA of 35 dB HL, the SRT was −13.0 dB SNR with a sensitivity of 88% and specificity of 93%. The normative binaural intelligibility level difference (BILD) value was 8.6 dB (SD = 2.0 dB) with normal-hearing participants. The learning effect due to the task and interface was 1.7 dB (1st to 7th test) and test duration was 3 min.ConclusionThe SoNoise test in its synthetic dichotic antiphasic presentation is a fast and reliable tool to diagnose hearing impairment at 20, 30, and 35 dB HL cut-offs.

An Optimized Sequence for Sparse Channel Estimation in a 5G MIMO System

ABSTRACT Recently, the massive Multiple-Input Multiple-Output (MIMO) system has been integrated with machine learning approaches to realise automatic channel state information detection. These methods need high computational complexity due to the lack of optimal training sequences and the failure to control the sparsity assembly based on massive MIMO. Moreover, it does not consider optimised training sequences based on a compressive sensing approach in the presence of contamination. Thus, this paper proposed an enhanced deep learning model with an optimised pilot training sequence for sparse channel estimation in a 5 G massive MIMO system. Initially, the system and sparse block channel impulse response model will be constructed by considering the time-domain synchronous Generalized Frequency Division Multiplexing (GFDM) system and additive white Gaussian noise (AWGN). In the proposed work, a seagull optimisation algorithm is developed to design the training pilot sequence depending on the coherence properties of the sensing matrix, which is used to recover the channel impulse response. Then, the sparse massive MIMO channel is evaluated by proposing a new deep channel estimator with an enhanced stacked auto encoder (D-ESAE). The proposed channel MSE is 10−6, the bit error rate is 0.04, and NMSE is 10−9 at 40 dB SNR through Matlab simulation.

Spread spectrum OFDM-IM with X-transform

This paper proposes a developed index modulated spread spectrum orthogonal frequency division multiplexing (OFDM) that is based on X-transform (IM-XOFDM-SS). The presented OFDM sends information not only by M-ary data symbols but also by indexes bits that used to select the spreading code from a known set, and applies the X-transform for baseband processing. A theoretical analysis for upper bound bit-error rate (BER) over Rayleigh fading channels is derived for the proposed scheme. Simulations results show that our proposed scheme outperforms the competitive benchmark systems in terms of BER, peak-to-average power ratio (PAPR) and computational complexity. It achieves, at 10−4 BER, about 22 dB and 21 dB SNR gain when spectral efficiency is SE=1 bps/Hz. Additionally, it gains about 5.5 dB PAPR reduction in comparison to IM-OFDM-SS scheme that is based on FFT for baseband processing.