Speaker Recognition Performance Research Articles

MULTI-CLASS AUTOMATED SPEECH LANGUAGE RECOGNITION USING NATURAL LANGUAGE PROCESSING WITH OPTIMAL DEEP LEARNING MODEL

With technological development, human–computer interaction (HCI) has improved, and spoken communication among machines and humans is one solution to enhance and expedite this process. Researchers have recently explored several systems to improve speech and speaker recognition performance in recent decades. A crucial threat in HCI is developing models that can effectually listen and respond like humans. It resulted in the development of the automated speech emotion recognition (SER) method, which can recognize various emotional classes by electing and extracting effectual features from speech signals. The fundamental problem of automated speech detection is the considerable variation in speech signals because of distinct speakers, language differences, speech differences, contents and acoustic conditions, voice modulation differences based on age and gender. With enhancements in deep learning (DL) and the affordability of computational resources, specifically graphical processing units (GPUs), research underwent a paradigm shift. Therefore, this study develops a multi-class automated speech language recognition using natural language processing with optimal deep learning (MASLR-NLPODL) technique. The MASLR-NLPODL technique intends to accomplish the efficient identification of different spoken languages. In the MASLR-NLPODL technique, the initial preprocessing technique involves windowing, frame blocking, and pre-emphasis block. Next, an adaptive time-frequency feature extractor approach utilizing the discrete fractional Fourier transform (DFrFT) was applied, which can be attained by extending the discrete Fourier transform (DFT) with eigenvectors. An improved Harris hawks optimization (IHHO) technique can be employed to select effectual features. Moreover, the classification of spoken languages can be performed by the gated recurrent unit (GRU) model. Finally, the salp swarm algorithm (SSA)-based hyperparameter selection process is involved in enhancing the performance of the GRU model. The design of the IHHO-based feature selection and SSA-based hyperparameter tuning process demonstrates the novelty of the work. The performance evaluation of the MASLR-NLPODL technique takes place under the VoxForge Dataset. The experimental validation of the MASLR-NLPODL technique exhibited a superior accuracy outcome of 96.40% over existing techniques.

Construction of an effective telephone bandwidth specifically for speaker recognition  

Although the telephone bandwidth is 0-4 kHz, the speaker specific information is not evenly distributed within this range but extends beyond 4 kHz. This is because the hypopharynx, especially the piriform fossa, affects the higher frequency region and contributes more to inter-speaker variation. By effectively shifting the higher bands to a lower frequency region, where speaker specific information is reduced, an effective 4 kHz bandwidth can be constructed to enhance speaker recognition performance. To achieve this a method was already proposed, which is extended in this paper to experimentally demonstrate and validate with more experiments. Furthermore, this paper defines the theoretically possible frequency space for which the frequency shifting method can be applied. To validate the method for different combinations of bands, possible bands were shifted in various directions in small steps. Speaker recognition experiments were conducted at each step to compare the performance against the baseband without any frequency shifting. Using the results of these extensive experiments, an approximate frequency space was defined where this frequency shifting performed better than the conventional baseband of 0-4 kHz signal. A simplified frequency shifting method was also investigated. Finally, the speech intelligibility of the frequency shifted narrow band speech signal was analysed using objective speech quality measures. This showed that intelligibility was not significantly affected by the frequency shifting method.

Enhancing Speaker Recognition Models with Noise-Resilient Feature Optimization Strategies

This paper delves into an in-depth exploration of speaker recognition methodologies, with a primary focus on three pivotal approaches: feature-level fusion, dimension reduction employing principal component analysis (PCA) and independent component analysis (ICA), and feature optimization through a genetic algorithm (GA) and the marine predator algorithm (MPA). This study conducts comprehensive experiments across diverse speech datasets characterized by varying noise levels and speaker counts. Impressively, the research yields exceptional results across different datasets and classifiers. For instance, on the TIMIT babble noise dataset (120 speakers), feature fusion achieves a remarkable speaker identification accuracy of 92.7%, while various feature optimization techniques combined with K nearest neighbor (KNN) and linear discriminant (LD) classifiers result in a speaker verification equal error rate (SV EER) of 0.7%. Notably, this study achieves a speaker identification accuracy of 93.5% and SV EER of 0.13% on the TIMIT babble noise dataset (630 speakers) using a KNN classifier with feature optimization. On the TIMIT white noise dataset (120 and 630 speakers), speaker identification accuracies of 93.3% and 83.5%, along with SV EER values of 0.58% and 0.13%, respectively, were attained utilizing PCA dimension reduction and feature optimization techniques (PCA-MPA) with KNN classifiers. Furthermore, on the voxceleb1 dataset, PCA-MPA feature optimization with KNN classifiers achieves a speaker identification accuracy of 95.2% and an SV EER of 1.8%. These findings underscore the significant enhancement in computational speed and speaker recognition performance facilitated by feature optimization strategies.

Analysis of Human Voice for Speaker Recognition: Concepts and Advancement

Human voice or speech is a contactless, non-invasive biometric trait for human recognition, easy to use with minimal computer complexity and inexpensive to implement. Speaker recognition (SR) has turned out to be a magnificent approach using speech as the central premise since decades. Its broad range of usages, like forensic speech verification to identify culprits by law enforcement authorities and access control to mobile banking, mobile shopping, etc., has made it a lucrative area of research. Also, the ease of use and dependability of SR will significantly assist people with disabilities in securely accessing and reaping the benefits of digital-era services. Additionally, the emergence of numerous deep learning methods for feature extraction and classification, has helped SR to achieve tremendous progress. This paper presents a comprehensive study on the progression of SR for decades till the present, including integration with Blockchain and challenges. It covers most of the factors that influence SR performance such as fundamentals and structure of SR, different speech pre-processing techniques, various speech features, feature extraction techniques, traditional and neural network-based classification techniques and deep learning-based SR toolkits. As a consequence, in this digital Blockchain era, it will help to design robust and reliable recognition-based services for mankind.

Speaker recognition using isomorphic graph attention network based pooling on self-supervised representation

The emergence of self-supervised representation (i.e., wav2vec 2.0) allows speaker-recognition approaches to process spoken signals through foundation models built on speech data. Nevertheless, effective fusion on the representation requires further investigating, due to the inclusion of fixed or sub-optimal temporal pooling strategies. Despite of improved strategies considering graph learning and graph attention factors, non-injective aggregation still exists in the approaches, which may influence the performance for speaker recognition. In this regard, we propose a speaker recognition approach using Isomorphic Graph ATtention network (IsoGAT) on self-supervised representation. The proposed approach contains three modules of representation learning, graph attention, and aggregation, jointly considering learning on the self-supervised representation and the IsoGAT. Then, we perform experiments for speaker recognition tasks on VoxCeleb1&2 datasets, with the corresponding experimental results demonstrating the recognition performance for the proposed approach, compared with existing pooling approaches on the self-supervised representation.

Exploring the performance of automatic speaker recognition using twin speech and deep learning-based artificial neural networks.

This study assessed the influence of speaker similarity and sample length on the performance of an automatic speaker recognition (ASR) system utilizing the SpeechBrain toolkit. The dataset comprised recordings from 20 male identical twin speakers engaged in spontaneous dialogues and interviews. Performance evaluations involved comparing identical twins, all speakers in the dataset (including twin pairs), and all speakers excluding twin pairs. Speech samples, ranging from 5 to 30 s, underwent assessment based on equal error rates (EER) and Log cost-likelihood ratios (Cllr). Results highlight the substantial challenge posed by identical twins to the ASR system, leading to a decrease in overall speaker recognition accuracy. Furthermore, analyses based on longer speech samples outperformed those using shorter samples. As sample size increased, standard deviation values for both intra and inter-speaker similarity scores decreased, indicating reduced variability in estimating speaker similarity/dissimilarity levels in longer speech stretches compared to shorter ones. The study also uncovered varying degrees of likeness among identical twins, with certain pairs presenting a greater challenge for ASR systems. These outcomes align with prior research and are discussed within the context of relevant literature.

Advancements in Speaker Recognition: Exploring Mel Frequency Cepstral Coefficients (MFCC) for Enhanced Performance in Speaker Recognition

Abstract: Speaker recognition, a fundamental capability of software or hardware systems, involves receiving speech signals, identifying the speaker present in the speech signal, and subsequently recognizing the speaker for future interactions. This process emulates the cognitive task performed by the human brain. At its core, speaker recognition begins with speech as the input to the system. Various techniques have been developed for speech recognition, including Mel frequency cepstral coefficients (MFCC), Linear Prediction Coefficients (LPC), Linear Prediction Cepstral coefficients (LPCC), Line Spectral Frequencies (LSF), Discrete Wavelet Transform (DWT), and Perceptual Linear Prediction (PLP). Although LPC and several other techniques have been explored, they are often deemed impractical for real-time applications. In contrast, MFCC stands out as one of the most prominent and widely used techniques for speaker recognition. The utilization of cepstrum allows for the computation of resemblance between two cepstral feature vectors, making it an effective tool in this domain. In comparison to LPC-derived cepstrum features, the use of MFCC features has demonstrated superior performance in metrics such as False Acceptance Rate (FAR) and False Rejection Rate (FRR) for speaker recognition systems. MFCCs leverage the human ear's critical bandwidth fluctuations with respect to frequency. To capture phonetically important characteristics of speech signals, filters are linearly separated at low frequencies and logarithmically separated at high frequencies. This design choice is central to the effectiveness of the MFCC technique. The primary objective of the proposed work is to devise efficient techniques that extract pertinent information related to the speaker, thereby enhancing the overall performance of the speaker recognition system. By optimizing feature extraction methods, this research aims to contribute to the advancement of speaker recognition technology.

Speaker Recognition Based on the Joint Loss Function

The statistical pyramid dense time-delay neural network (SPD-TDNN) model makes it difficult to deal with the imbalance of training data, poses a high risk of overfitting, and has weak generalization ability. To solve these problems, we propose a method based on the joint loss function and improved statistical pyramid dense time-delay neural network (JLF-ISPD-TDNN), which improves on the SPD-TDNN model and uses the joint loss function method to combine the advantages of the cross-entropy loss function and the comparative learning of the loss function. By minimizing the distance between speech embeddings from the same speaker and maximizing the distance between speech embeddings from different speakers, the model could achieve enhanced generalization performance and more robust speaker feature representation. We evaluated the proposed method’s performance using the evaluation indexes of the equal error rate (EER) and minimum cost function (minDCF). The experimental results show that the EEE and minDCF on the Aishell-1 dataset reached 1.02% and 0.1221%, respectively. Therefore, using the joint loss function in the improved SPD-TDNN model can significantly enhance the model’s speaker recognition performance.

Joint filter combination-based central difference feature extraction and attention-enhanced Dense-Res2Block network for short-utterance speaker recognition

Speech authentication in smart services typically involves short utterances. However, due to the short duration of these utterances (e.g. less than 3 s) and the limited enrolment and/or test data available, it is difficult to learn enough information to accurately distinguish the person. As a result, speaker recognition from short utterances is very challenging. In this paper, in the acoustic end. we propose a novel Bark-scaled Gaussian and linear filter bank cepstral coefficients (BGLCC) and multi-dimensional central difference (MDCD) acoustic features extraction method. Also, to enhance the discriminative embedding, in the network end, a novel attention-enhanced Dense-Res2Block network is proposed. First, the rich low-frequency information is extracted based on the high distribution density of the Bark-scaled Gaussian filter bank in the low-frequency domain, and more high-frequency information is extracted based on the linear filter bank uniformly distributed in the high-frequency domain. This means that the combined filters can produce more discriminative and richer acoustic features from short-duration audio signals. In addition, the multi-dimensional central difference method captures better speaker dynamic features in the relative BGLCC domain to improve the performance of short utterance speaker recognition. Finally, an attention-enhanced Dense-Res2Block architecture can obtain a variety of feature expressions of different scale combinations to enhance features. Extensive analysis of a variety of datasets, which speech samples of different types, diverse lengths, etc., demonstrate the superiority of the proposed feature extraction method and model over existing acoustic feature extraction methods and speaker recognition models, including those based on MFCCs, LPCCs, and fusion features, and X-vector-PLDA, BLSTM-ResNet, ResNet34-SP, ECAPA-TDNN respectively. The experimental results show that the proposed method achieves the best performance compared to the existing approach.

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

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

Combined Bidirectional Long Short-Term Memory with Mel-Frequency Cepstral Coefficients Using Autoencoder for Speaker Recognition

Recently, neural network technology has shown remarkable progress in speech recognition, including word classification, emotion recognition, and identity recognition. This paper introduces three novel speaker recognition methods to improve accuracy. The first method, called long short-term memory with mel-frequency cepstral coefficients for triplet loss (LSTM-MFCC-TL), utilizes MFCC as input features for the LSTM model and incorporates triplet loss and cluster training for effective training. The second method, bidirectional long short-term memory with mel-frequency cepstral coefficients for triplet loss (BLSTM-MFCC-TL), enhances speaker recognition accuracy by employing a bidirectional LSTM model. The third method, bidirectional long short-term memory with mel-frequency cepstral coefficients and autoencoder features for triplet loss (BLSTM-MFCCAE-TL), utilizes an autoencoder to extract additional AE features, which are then concatenated with MFCC and fed into the BLSTM model. The results showed that the performance of the BLSTM model was superior to the LSTM model, and the method of adding AE features achieved the best learning effect. Moreover, the proposed methods exhibit faster computation times compared to the reference GMM-HMM model. Therefore, utilizing pre-trained autoencoders for speaker encoding and obtaining AE features can significantly enhance the learning performance of speaker recognition. Additionally, it also offers faster computation time compared to traditional methods.

Speakers are more cooperative and less individual when interacting in larger group sizes.

Cooperation, acoustically signaled through vocal convergence, is facilitated when group members are more similar. Excessive vocal convergence may, however, weaken individual recognizability. This study aimed to explore whether constraints to convergence can arise in circumstances where interlocutors need to enhance their vocal individuality. Therefore, we tested the effects of group size (3 and 5 interactants) on vocal convergence and individualization in a social communication scenario in which individual recognition by voice is at stake. In an interactive game, players had to recognize each other through their voices while solving a cooperative task online. The vocal similarity was quantified through similarities in speaker i-vectors obtained through probabilistic linear discriminant analysis (PLDA). Speaker recognition performance was measured through the system Equal Error Rate (EER). Vocal similarity between-speakers increased with a larger group size which indicates a higher cooperative vocal behavior. At the same time, there was an increase in EER for the same speakers between the smaller and the larger group size, meaning a decrease in overall recognition performance. The decrease in vocal individualization in the larger group size suggests that ingroup cooperation and social cohesion conveyed through acoustic convergence have priority over individualization in larger groups of unacquainted speakers.

Familiar and unfamiliar speaker recognition assessment and system emulation for cochlear implant users.

In the area of speech processing, human speaker identification under naturalistic environments is a challenging task, especially for hearing-impaired individuals with cochlear implants (CIs) or hearing aids (HAs). Motivated by the fact that electrodograms reflect direct CI stimulation of input audio, this study proposes a speaker identification (ID) investigation using two-dimensional electrodograms constructed from the responses of a CI auditory system to emulate CI speaker ID capabilities. Features are extracted from electrodograms through an identity vector (i-vector) framework to train and generate identity models for each speaker using a Gaussian mixture model-universal background model followed by probabilistic linear discriminant analysis. To validate the proposed system, perceptual speaker ID for 20 normal hearing (NH) and seven CI listeners was evaluated with a total of 41 different speakers and compared with the scores from the proposed system. A one-way analysis of variance showed that the proposed system can reliably predict the speaker ID capability of CI (F[1,10] = 0.18, p = 0.68) and NH (F[1,20] = 0, p = 0.98) listeners in naturalistic environments. The impact of speaker familiarity is also addressed, and the results show a reduced performance for speaker recognition by CI subjects using their CI processor, highlighting limitations of current speech processing strategies used in CIs/HAs.

Vowel convergence does not affect auditory speaker discriminability in humans and machine in a case study on Swiss German dialects

In this study, we examined whether the convergence in interlocutors’ vowel acoustics leads to decreasing discriminability between interlocutors’ voices. Ten pairs of Grison and Zürich German speakers produced lexical items before and after dialogue interactions with evidence of vowel convergence in post-dialogue productions. In Experiment 1, native and non-native Swiss German listeners discriminated pairs of speakers whose speech was obtained pre- and post-dialogue. Results showed that listeners’ sensitivity (A’) was higher for native than non-native listeners, but comparable for pre- and post-dialogue recordings. The observed negative correlation between voice discrimination and the acoustic distance in formant space was mainly driven by a single speaker pair. In Experiment 2, the speaker recognition performance of an i-vector-based software was compared in pre- and post-dialogue speech. Results revealed no difference in the system performance between the two conditions. The findings suggest that vowel convergence does not compromise voice discriminability under the given experimental conditions.

Fast and Accurate Voice Biometrics with Deep Learning Algorithm of CNN Depthwise Separable Convolution Model and Fusion of DWT-MFCC Methods

Theft of private data became a threat of crime in cyberspace. This issue was in line with rapid development of data technology, especially online transactions. To attenuate this problem, voice biometrics was developed as an answer to keep up security identity. This paper develops the voice biometric framework based on Convolutional Neural Network Depthwise Separable Convolution (DSC) model and the fusion of Discrete Wavelet Transform (DWT) and Mel Frequency Cepstral Coefficients (MFCC). Such a scheme has targeted to increase the high accuracy, to reduce the burden of high computational costs and to speed up the performance of classification process time. We conduct three testing performance, i.e. voice Biometric Training Performance, speaker Recognition Performance (”Who is speaking?”), and Speech Recognition performance (”What keyword is uttered?”). For each of the testing, the results are compared with CNN Standard performance. The training results has shown that CNN DSC model has reduced the amount of training parameters to 364,506, leading to accelerate the performance of training process time to 5.12 minutes. The results of speaker recognition performance has attained the best performance with an accuracy 99.25%, precision 97.14%, recall 98.17% and F1-score 97.28%. The results of speech recognition performance has been able to improve the best performance with accuracy 100%. It can be concluded that CNN DSC with the fusion of DWT- MFCC has outperformed the CNN Standard. The framework can be applied for the identification and verification of user voices accurately, quickly and efficiently for any applications requiring better security performance.

Multi-Source Domain Adaptation for Text-Independent Forensic Speaker Recognition

Adapting speaker recognition systems to new environments is a widely-used technique to improve a well-performing model learned from large-scale data towards a task-specific small-scale data scenarios. However, previous studies focus on single domain adaptation, which neglects a more practical scenario where training data are collected from multiple acoustic domains needed in forensic scenarios. Audio analysis for forensic speaker recognition offers unique challenges in model training with multi-domain training data due to location/scenario uncertainty and diversity mismatch between reference and naturalistic field recordings. It is also difficult to directly employ small-scale domain-specific data to train complex neural network architectures due to domain mismatch and performance loss. Fine-tuning is a commonly-used method for adaptation in order to retrain the model with weights initialized from a well-trained model. Alternatively, in this study, three novel adaptation methods based on domain adversarial training, discrepancy minimization, and moment-matching approaches are proposed to further promote adaptation performance across multiple acoustic domains. A comprehensive set of experiments are conducted to demonstrate that: 1) diverse acoustic environments do impact speaker recognition performance, which could advance research in audio forensics, 2) domain adversarial training learns the discriminative features which are also invariant to shifts between domains, 3) discrepancy-minimizing adaptation achieves effective performance simultaneously across multiple acoustic domains, and 4) moment-matching adaptation along with dynamic distribution alignment also significantly promotes speaker recognition performance on each domain, especially for the LENA-field domain with noise compared to all other systems.

Lightweight Speaker Recognition in Poincaré Spaces

This letter proposes a lightweight model for speaker recognition by leveraging a hyperbolic space. The speaker recognition performance heavily depends on the distinctiveness of speaker embeddings induced by metric learning. However, most state-of-the-art embedding methods are typically based on the Euclidean metric space, which does not account for inherent hierarchical structures of speech voice characteristics. The recent development of the neural hyperbolic geometry has demonstrated its effectiveness to model continuous hierarchical structures, which have been typically cumbersome to model by standard deep neural networks. This facet provides an additional by-product of a compact representation. Inspired by the favorable geometry of the hyperbolic geometry, we developed a hyperbolic ResNet for speaker recognition. We found that in smaller dimension regimes than typical cases, the learned speaker embeddings are more discriminative; in other words, more compact at the same level of performance. Our experiments on the large-scale VoxCeleb datasets show that, given the limited channel dimensions of neural networks, our method consistently has favorable performance against the standard ResNet for both speaker recognition and verification tasks.

When Speaker Recognition Meets Noisy Labels: Optimizations for Front-Ends and Back-Ends

A typical speaker recognition system often involves two modules: a feature extractor front-end and a speaker identification back-end. Despite the superior performance that deep neural networks have achieved for the front-end, their success benefits from the availability of large-scale, correctly labeled datasets. While label noise is unavoidable in speaker recognition datasets, both the front-end and back-end are affected by label noise, which degrades speaker recognition performance. In this paper, we first conduct comprehensive experiments to help improve our understanding of the effects of label noise on both the front-end and back-end. Then, we propose a simple yet effective training paradigm and loss correction method to handle label noise in the front-end. We combine our proposed method with the recently proposed Bayesian estimation of PLDA for noisy labels, and the whole system shows strong robustness to label noise. Furthermore, we show two practical applications of the improved system: one application corrects noisy labels based on an utterance’s chunk-level predictions, and the other algorithmically filters out high-confidence noisy samples within a dataset. By applying the second application to the NIST SRE04–10 dataset and verifying filtered utterances by human validation, we identify that approximately 1% of the NIST SRE04–10 dataset is made up of label errors.

Multi-lingual Speaker Recognition based on Asymmetric Convolution and Central Loss Function

In order to solve the problem of poor speaker recognition performance on multi-language corpus on convolutional neural networks and large amount of calculations on multiple parameters, this paper draws on asymmetric convolution and center loss (Center Loss) functions to improve and optimize the ResNet model. And perform speaker recognition tasks in the Chinese-Uyghur corpus. The results show that, compared with the original model, the improved model has higher accuracy, fewer parameters, and reduced computational complexity.

Nonlinear waveform distortion: Assessment and detection of clipping on speech data and systems

Speech, speaker, and language systems have traditionally relied on carefully collected speech material for training acoustic models. There is an enormous amount of freely accessible audio content. A major challenge, however, is that such data is not professionally recorded, and therefore may contain a wide diversity of background noise, nonlinear distortions, or other unknown environmental or technology-based contamination or mismatch. There is a crucial need for automatic analysis to screen such unknown datasets before acoustic model development training, or to perform input audio purity screening prior to classification. In this study, we propose a waveform based clipping detection algorithm for naturalistic audio streams and examine the impact of clipping at different severities on speech quality measurements and automatic speaker recognition systems. We use the TIMIT and NIST SRE08 corpora as case studies. The results show, as expected, that clipping introduces a nonlinear distortion into clean speech data, which reduces speech quality and performance for speaker recognition. We also investigate what degree of clipping can be present to sustain effective speech system performance. The proposed detection system, which will be released, could contribute to massive new audio collections for speech and language technology development (e.g. Google Audioset (Gemmeke et al., 2017), CRSS-UTDallas Apollo Fearless-Steps (Yu et al., 2014) (19,000 h naturalistic audio from NASA Apollo missions)).