A5053479846- Speech and Audio Processing
- Advanced Adaptive Filtering Techniques
- Speech Recognition and Synthesis
- Indoor and Outdoor Localization Technologies
- Blind Source Separation Techniques
- Digital Media Forensic Detection
- GNSS positioning and interference
- Hearing Loss and Rehabilitation
- Seismology and Earthquake Studies
- Chaos-based Image/Signal Encryption
- Music and Audio Processing
- Advanced Steganography and Watermarking Techniques
Microsoft (United States)
2024
Ruhr University Bochum
2015-2019
The ICASSP 2023 Deep Noise Suppression (DNS) Challenge marks the fifth edition of DNS challenge series. challenges were organized from 2019 to foster research in field DNS. Previous held at INTERSPEECH 2020, 2021, and 2022. This aims advance models capable jointly addressing denoising, dereverberation, interfering talker suppression, with separate tracks focusing on headset speakerphone scenarios. facilitates personalized deep noise suppression by providing accompanying enrollment clips for...
In this paper, we present and compare novel algorithms to localize simultaneous speakers using four microphones distributed on a pair of binaural hearing aids. The framework consists two groups localization algorithms, namely, beamforming-based statistical model based algorithms. We first generalize our previously proposed methods beamforming techniques the configuration with 2 × microphones. Next, contribute for maximum likelihood approach that also takes head-related transfer functions...
In this paper we present a novel algorithm to localize and separate simultaneous speakers using hearing aids when the head is subject rotational movement. Most of algorithms used in are able extract target signals that look direction user suffer from reduced performance localizing sounds received other directions. Moreover, head-shadowing as well variations like movements may lead significant distortions. The proposed binaural GSC beamformer includes an MMSE-based localization ITD/ILD model...
This article addresses the problem of distance estimation using binaural hearing aid microphones in reverberant rooms. Among several indicators, direct-to-reverberant energy ratio (DRR) has been shown to be more effective than other features. Therefore, we present two novel approaches estimate DRR signals. The first method is based on interaural magnitude-squared coherence whereas second approach uses stochastic maximum likelihood beamforming power direct and components. proposed algorithms...
In this paper we investigate two methods for the preservation of spatial cues in binaural speaker separation. We develop these as extensions our previously proposed model-based generalized sidelobe canceller (GSC) which utilizes a maximum likelihood technique localization. implementation adaptive GSC provides an estimate target signal well estimation presence probability (TPP). Binaural outputs are generated different ways: first approach signals rendered using output combined with HRTF...
In this paper we propose a novel multi-channel algorithm to separate simultaneous speakers in an environment where the microphone array is subject movement. When microphones are mounted person's head, for instance, movements can lead ambiguities with respect sources and distortions processed signal. The proposed system estimates direction-of-arrival of speaker's signals relative updates these using inertial measurement unit (IMU). A GMM-based localization model used compute posterior...
In this work we evaluate the effects of head radius on binaural localization algorithms. We employ a spherical model and null-steering beamforming method. The characterizes cues in form HRTFs. One main parameters is radius. propose to optimize jointly for both source location contrast free-field configuration where it difficult estimate microphone distance simultaneously, algorithm yields unique solution Moreover, real recordings show that commonly-assumed size achieves fairly reliable...
This paper proposes a novel algorithm to estimate the direct-to-reverberant energy ratio (DRR) using hearing aid microphones. The is based on interaural magnitude-squared coherence of signals and able take both phase level differences microphones in binaural configuration into account. We employ spherical head model approximate cues. proposed uses common assumption an ideally diffuse reverberation sound field. test our approach simulated measured room impulse responses. Results show improved...
A new steganalysis system for JPG-based image data hiding is proposed in this paper. We use features extracted from both wavelet and DCT domains that are refined later the sense of utmost discrimination between clear stego images classification system. Statistical properties SVD sub-bands combined with extended DCT-Markov features, most sensitive to embedding chosen through a SVM-RFE based selection algorithm. Experimental results show significant improvement over baseline methods,...
Deep Speech Enhancement Challenge is the 5th edition of deep noise suppression (DNS) challenges organized at ICASSP 2023 Signal Processing Grand Challenges. DNS were during 2019-2023 to stimulate research in speech enhancement (DSE). Previous INTERSPEECH 2020, 2021, and 2022. From prior editions, we learnt that improving signal quality (SIG) challenging particularly presence simultaneously active interfering talkers noise. This challenge aims develop models for joint denosing,...