Enhanced Forensic Speaker Verification Using a Combination of DWT and MFCC Feature Warping in the Presence of Noise and Reverberation Conditions

Al-Ali, Ahmed Kamil Hasan; Dean, David; Senadji, Bouchra; Chandran, Vinod; Naik, Ganesh R.

doi:10.1109/access.2017.2728801

Cited by 64 publications

(40 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Besides, the discrete wavelet transform (DWT) is a prevalent tool for analyzing the acoustic signal in time and frequency. Many researchers have introduced the wavelet transform (WT) into the extraction of feature warped MFCC (FW-MFCC), for instance, Ahmed Kamil Hasan et al combine FW-MFCC and DWT for speaker verification [37], [38]. They apply DWT to decompose the speech into the low frequency sub-band and the high frequency sub-band coefficients, and then splice the frequency response of the wavelet coefficients into a complete frequency spectrum, finally, they acquire the Mel logarithmic power spectrum by calculating the wavelet coefficient energy.…”

Section: ) Feature Warped Mfcc and Dwtmentioning

confidence: 99%

HCI on the Table: Robust Gesture Recognition Using Acoustic Sensing in Your Hand

et al. 2020

View full text Add to dashboard Cite

The paper proposes a new HCI mechanism for device-free gesture recognition on the table using acoustic signal, which can extend the gesture input and interactions beyond the tiny screen of mobile device and allow users to provide input without blocking screen view. Previous researches have either relied on additional devices (e.g., special wearable device and mouse) or required active acoustic signals which demand additional cost and be less prone to popularize, while we explore the device-free gesture recognition using passive acoustic signals. This technology is more challenging due to the lack of an effective approach to eliminate the inherent ambient noise disturbances and extract stable gesture features. We fuse both short time energy (STE) and zero-crossing rate (ZCR) to identify the effective signals from the original input, and leverage the Mel frequency cepstral coefficients (MFCC), cochlear filter cepstral coefficients (CFCC) to extract the stable features from different gestures. The unique features in support vector machine (SVM) classifier achieve a high gesture recognition accuracy from the noisy scenarios and mismatched conditions. Implementation on the Android system has realized real-time processing of the feature extraction and gesture recognition. Extensive evaluations show our algorithm has a better noise tolerant performance and the system could recognize seven common gestures (click, flip left/right, scroll up/down, zoom in/out) on smart devices with an accuracy of 93.2%.

show abstract

Section: ) Feature Warped Mfcc and Dwtmentioning

confidence: 99%

HCI on the Table: Robust Gesture Recognition Using Acoustic Sensing in Your Hand

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Алгоритм MFCC часто встречается в предлагаемых научным сообществом автоматических системах распознавания речи [28][29][30][31][32]. Тем не менее процесс параметризации нередко включает в себя не один метод, а комбинацию из нескольких [33][34][35][36][37]. Большое разнообразие сочетаний подходов можно наблюдать у участников ASVspoof: Automatic Speaker Verification Spoofing and Countermeasures Challenge [38], решающих задачу противодействия спуфинговым атакам.…”

Section: заключениеunclassified

Overview of methods for extracting acoustic speech features in speaker recognition

Sudjenkova¹

2020

Transaction of Scientific Papers of the NSTU

View full text Add to dashboard Cite

Биометрические технологии являются перспективным направлением в области информационной безопасности. Голосовая биометрия на сегодняшний день широко распространена, и работы над повышением качества голосовых систем не теряют своей актуальности. Выбор метода извлечения речевых признаков-один из ключевых этапов проектирования голосовых автоматических систем. В статье рассматриваются акустические параметры, обусловленные физиологическими свойствами речевого тракта человека: частота основного тона, огибающая спектра, форманты и антиформанты. Тема статьи касается методов их извлечения. Большую часть составляют различные варианты кепстрального анализа, поскольку именно они наиболее часто встречаются в современных разработках как в виде использования популярных мел-частотных кепстральных коэффициентов, так и в новых модификациях. Также внимание уделяется алгоритмам линейного предсказания, спектрального центроида и вейвлет-анализа. Параметризация речевых характеристик входит в распознавание речи, эмоций, языка, гендера. Хотя в статье перечислены основные подходы извлечения акустических признаков речи с целью распознавания диктора, материал может быть полезен и в вышеперечисленных задачах обработки речевых сигналов Ключевые слова: распознавание диктора, анализ речи, кепстральные коэффициенты, линейное предсказание, перцептивное линейное предсказание, спектральный центроид, вейвлет-анализ ВВЕДЕНИЕ Биометрические технологии активно внедряются в жизнь общества. Об этом свидетельствует существующий и прогнозируемый рост рынка биометрии как на мировом, так и на отечественном уровне [1]. Распознавание по голосу благодаря широкой доступности оборудования, возможности удаленной идентификации, простому процессу обучения и ис-* Статья получена 26 сентября 2019 г.

show abstract

“…At the same time, the systems are expected to be robust to noisy and reverberant environments, where they are typically used. Recent studies have suggested speech enhancement (SE) algorithms to improve the robustness of the SV systems to noise and reverberation [25]- [29].…”

Section: Introductionmentioning

confidence: 99%

A Unified Deep Learning Framework for Short-Duration Speaker Verification in Adverse Environments

et al. 2020

View full text Add to dashboard Cite

Speaker verification (SV) has recently attracted considerable research interest due to the growing popularity of virtual assistants. At the same time, there is an increasing requirement for an SV system: it should be robust to short speech segments, especially in noisy and reverberant environments. In this paper, we consider one more important requirement for practical applications: the system should be robust to an audio stream containing long non-speech segments, where a voice activity detection (VAD) is not applied. To meet these two requirements, we introduce feature pyramid module (FPM)-based multiscale aggregation (MSA) and self-adaptive soft VAD (SAS-VAD). We present the FPM-based MSA to deal with short speech segments in noisy and reverberant environments. Also, we use the SAS-VAD to increase the robustness to long non-speech segments. To further improve the robustness to acoustic distortions (i.e., noise and reverberation), we apply a masking-based speech enhancement (SE) method. We combine SV, VAD, and SE models in a unified deep learning framework and jointly train the entire network in an endto-end manner. To the best of our knowledge, this is the first work combining these three models in a deep learning framework. We conduct experiments on Korean indoor (KID) and VoxCeleb datasets, which are corrupted by noise and reverberation. The results show that the proposed method is effective for SV in the challenging conditions and performs better than the baseline i-vector and deep speaker embedding systems.

show abstract

Enhanced Forensic Speaker Verification Using a Combination of DWT and MFCC Feature Warping in the Presence of Noise and Reverberation Conditions

Cited by 64 publications

References 31 publications

HCI on the Table: Robust Gesture Recognition Using Acoustic Sensing in Your Hand

HCI on the Table: Robust Gesture Recognition Using Acoustic Sensing in Your Hand

Overview of methods for extracting acoustic speech features in speaker recognition

A Unified Deep Learning Framework for Short-Duration Speaker Verification in Adverse Environments

Contact Info

Product

Resources

About