Hierarchical and Parallel Processing of Modulation Spectrum for ASR applications

Chat with Graph Search

Ask any question about EPFL courses, lectures, exercises, research, news, etc. or try the example questions below.

DISCLAIMER: The Graph Chatbot is not programmed to provide explicit or categorical answers to your questions. Rather, it transforms your questions into API requests that are distributed across the various IT services officially administered by EPFL. Its purpose is solely to collect and recommend relevant references to content that you can explore to help you answer your questions.

Ultra-low noise lasers are essential tools in various applications, including data communication, light detection and ranging (LiDAR), quantum computing and sensing, and optical metrology. Recent advances in integrated photonics, specifically the development of the ultra-low loss silicon nitride (

Si3N4

) platform, have allowed attaining performance that exceeds conventional legacy laser systems, including the phase noise of fiber lasers. This platform can be combined with the monolithic integration of piez

The increasing complexity of deep neural networks (DNNs) necessitates effective model compression to reduce their computational and memory footprints for deployment on resource-constrained hardware. Layer-wise bit allocation is a prominent compression method shown to significantly reduce DNN footprints while preserving model accuracy. However, how best to incorporate hardware constraints within the allocation search remains a key question, as many tacitly assume constraints can be adequately handled via sof

Conventionally, speech emotion recognition has been approached by utterance or turn-level modelling of input signals, either through extracting hand-crafted low-level descriptors, bag-of-audio-words features or by feeding long-duration signals directly to deep neural networks (DNNs). While this approach has been successful, there is a growing interest in modelling speech emotion information at the short segment level, at around 250ms-500ms (e.g. the 2021-22 MuSe Challenges). This paper investigates both han

Shaping of light behind masks using different techniques is the milestone of the printing industry. The aerial image distribution or the intensity distribution at the printing distances defines the resolution of the structure after printing. Contrast and phase are the two parameters that play a major role in shaping of light to get the desired intensity pattern. Here, in contrast to many other contributions that focus on intensity, we discuss the phase evolution for different structures. The amplitude or in

An extensive study of digital pulse processing methods is presented. Existing methods, both traditional and more recent, are compared with original advanced techniques within an appropriate modeling and benchmarking framework. This comprehensive approach e ...

A Bulk Acoustic Wave (BAW) resonator based 2.4-GHz low power receiver is presented in this work. The intrinsic high quality factor ( Q) of the BAW resonator (around 400 at 2.4-GHz) is exploited to provide channel selection at RF and in the frequency synthesis. A novel way of addressing multiple channels (arbitrary frequency) using integer dividers and a BAW digitally controlled oscillator (DCO) and thus avoiding the need for a PLL is proposed in this work. To the best of our knowledge, this work is the firs

This paper presents the OpenMP parallel implementation of the Distinct Lattice Spring Model (DLSM). The motivation of this study is to reduce the computing time and increase the computational capacity of the DLSM code. Firstly, the basic theory of DLSM is introduced. As an explicit method, the DLSM is very suitable for parallelization implementation. Only a few code changes are needed for the OpenMP implementation. Then; the detail parallel design of the OpenMP implementation is presented. Finally, numerica

In this paper, we consider an impulse-based ultrawideband (UWB) receiver and assess the performance of different multiple access and modulation schemes, including time-hopping pulse position modulation (TH-PPM), time-hopping binary phase shift keying (TH-BPSK), and BPSK-BPSK, under both the additive white Gaussian noise (AWGN) and indoor multipath (CM3) channel assumptions. The assumed receiver combines the received pulses to increase the signal to noise ratio (SNR) and correlates the summed signal with the

Automatic Speech Recognition (ASR) systems usually fail when they encounter speech from far-field microphone in reverberant environments. This is due to the application of short-term feature extraction techniques which do not compensate for the artifacts introduced by long room impulse responses. In this paper, we propose a front-end, based on Frequency Domain Linear Prediction (FDLP), that tries to remove reverberation artifacts present in far-field speech. Long temporal segments of far-field speech are an

Performance of a typical automatic speech recognition (ASR) system severely degrades when it encounters speech from reverberant environments. Part of the reason for this degradation is the feature extraction techniques that use analysis windows which are much shorter than typical room impulse responses. We present a feature extraction technique based on modeling temporal envelopes of the speech signal in narrow sub-bands using Frequency Domain Linear Prediction (FDLP). FDLP provides an all-pole approximatio

This paper investigates a neural network based acoustic feature mapping to extract robust features for automatic speech recognition (ASR) of overlapping speech. In our preliminary studies, we trained neural networks to learn the mapping from log mel filter bank energies (MFBEs) extracted from the distant microphone recordings, including multiple overlapping speakers, to log MFBEs extracted from the clean speech signal. In this paper, we explore the mapping of higher order mel-filterbank cepstral coefficient

An experimental method is presented for characterization of the combined intensity and frequency modulation produced when the injection current of a laser diode is modulated. The reported technique is based on the analysis of the harmonic signals produced when a modulated laser is used to probe a gas absorption line by the so-called wavelength-modulation spectroscopy method. Based on a theoretical model of this technique, we present two methods that facilitate the determination of (i) the deviation in laser

We address the problem of clustering multimodal group actions in meetings using a two-layer HMM framework. Meetings are structured as sequences of group actions. Our approach aims at creating one cluster for each group action, where the number of group actions and the action boundaries are unknown a priori. In our framework, the first layer models typical actions of individuals in meetings using supervised HMM learning and low-level audio-visual features. A number of options that explicitly model certain as

An electron cyclotron resonance heating (ECRH) system has been designed for JET in the framework of the JET enhanced performance project (JET-EP) under the European fusion development agreement. Due to financial constraints it has been decided not to implement this project. Nevertheless, the design work conducted from April 2000 to January 2002 shows a number of features that can be relevant in preparation of future ECRH systems, e.g. for ITER. The ECRH system was foreseen to comprise six gyrotrons, 1 MW ea

During the last decade the field of computational chem. has experienced an enormous progress. Due to the exponential increase in computer power and the development of new computational methods it has now become possible to treat many complex chem. and biol. problems on an accurate and realistic level. Ab initio mol. dynamics simulations based on d. functional theory are among the most powerful of these new tools. Using parallel computers, systems of a few hundred of atoms can be studied routinely. By extend

The temp. dependence of internucleotide nitrogen-nitrogen scalar couplings 2hJ(N,N) across hydrogen bonds in adenine-uracil (A-U) and guanine-cytosine (G-C) base pairs of the 22 nucleotide RNA oligomer GGCGAAGUCGAAAGAUGGCGCC was studied between 280 and 310 K. The value of 2hJ(N,N) was obsd. to decrease monotonically for all four base pairs with increasing temp. The temp. dependence of 2hJ(N,N) was found to be more pronounced for the A-U base pair than for G-C base pairs. An earlier study of cross-correlatio

In this paper, we present a new approach towards user-custom-ized password speaker verification combining the advantages of hybrid HMM/ANN systems, using Artificial Neural Networks (ANN) to estimate emission probabilities of Hidden Markov Models, and Gaussian Mixture Models. In the approach presented here, we indeed exploit the properties of hybrid HMM/ANN systems, usually resulting in high phonetic recognition rates, to automatically infer the baseline phonetic transcription (HMM topology) associated with

The multi-band processing paradigm for noise robust ASR was originally motivated by the observation that human recognition appears to be based on independent processing of separate frequency sub-bands, and also by

missing data'' results which have shown that ASR can be made significantly more robust to band-limited noise if noisy sub-bands can be detected and then ignored. Of the different multi-band models which have been proposed, only the

Full Combination'' or ``all-wise'' multi-band HMM/ANN hybrid a

Picosecond transient reflectometry and dynamic grating techniques have been applied to investigate the perpendicular and parallel transport of optically excited carriers in strained-layer Si0.83Ge0.17/Si superlattices. We present results of the carrier ambipolar diffusivity and effective lifetime measurements in the layered structure and substrate within the 10(17)-10(20) cm(-3) density range. The combined experimental data are discussed in terms of parallel and perpendicular diffusion of carriers, interfac

A solution to the full two-dimensional eigenvalue problem of electrostatic microinstabilities in a tokamak plasma is presented in the framework of gyrokinetic theory. The approach is the generalization of methods previously developed for a cylindrical system [S. Brunner and J. Vaclavik, Phys. Plasmas 5, 365 (1998)]. By solving the spectral problem in a special Fourier space adapted to the curved geometry, orbit width as well as Larmor radius can be kept to all orders. For a first numerical implementation, a

Hierarchical and Parallel Processing of Modulation Spectrum for ASR applications

Chat with Graph Search

Monolithic piezoelectrically tunable hybrid integrated laser with sub-fiber laser coherence

Constrained bit allocation for neural networks

Towards Learning Emotion Information From Short Segments of Speech

Phase analysis of amplitude binary mask structures

Comparative analysis of digital pulse processing methods at high count rates

A 2.4-GHz MEMS-Based PLL-Free Multi-Channel Receiver With Channel Filtering at RF

OpenMP parallelization of the three-dimensional distinct lattice spring model

Performance Comparison of UWB Impulse-Based Multiple Access Schemes in Indoor Multipath Channels

Front-end for Far-field Speech Recognition based on Frequency Domain Linear Prediction

Recognition Of Reverberant Speech Using Frequency Domain Linear Prediction

Neural Network based Regression for Robust Overlapping Speech Recognition using Microphone Arrays

Experimental method based on wavelength-modulation spectroscopy for the characterization of semiconductor lasers under direct modulation

Multimodal Group Action Clustering in Meetings

The design of an ECRH system for JET-EP

Mixed quantum mechanical/molecular mechanical simulation of biological systems

Temperature dependence of internucleotide nitrogen-nitrogen scalar couplings in RNA

User-Customized Password Speaker Verification based on HMM/ANN and GMM Models

From Multi-Band Full Combination to Multi-Stream Full Combination Processing in Robust ASR

Lateral and cross-well transport of highly and moderately excited carriers in Si1-xGex/Si superlattices

Global approach to the spectral problem of microinstabilities in tokamak plasmas using a gyrokinetic model