scholarly journals On Training Targets for Supervised LPC Estimation to Augmented Kalman Filter-based Speech Enhancement

2021 ◽  
Author(s):  
Sujan Kumar Roy ◽  
Aaron Nicolson ◽  
Kuldip K. Paliwal
Keyword(s):  
Power Spectrum ◽  
Speech Enhancement ◽  
Linear Prediction ◽  
Real Life ◽  
Large Data ◽  
Convolutional Network ◽  
Data Set ◽  
Listening Tests ◽  
Objective Quality ◽  
Prediction Coefficient

The performance of speech coding, speech recognition, and speech enhancement largely depends upon the accuracy of the linear prediction coefficient (LPC) of clean speech and noise in practice. Formulation of speech and noise LPC estimation as a supervised learning problem has shown considerable promise. In its simplest form, a supervised technique, typically a deep neural network (DNN) is trained to learn a mapping from noisy speech features to clean speech and noise LPCs. Training targets for DNN to clean speech and noise LPC estimation fall into four categories: line spectrum frequency (LSF), LPC power spectrum (LPC-PS), power spectrum (PS), and magnitude spectrum (MS). The choice of appropriate training target as well as the DNN method can have a significant impact on LPC estimation in practice. Motivated by this, we perform a comprehensive study on the training targets using two state-of-the-art DNN methods--- residual network and temporal convolutional network (ResNet-TCN) and multi-head attention network (MHANet). This study aims to determine which training target as well as DNN method produces more accurate LPCs in practice. We train the ResNet-TCN and MHANet for each training target with a large data set. Experiments on the NOIZEUS corpus demonstrate that the LPC-PS training target with MHANet produces a lower spectral distortion (SD) level in the estimated speech LPCs in real-life noise conditions. We also construct the AKF with the estimated speech and noise LPC parameters from each training target using ResNet-TCN and MHANet. Subjective AB listening tests and seven different objective quality and intelligibility evaluation measures (CSIG, CBAK, COVL, PESQ, STOI, SegSNR, and SI-SDR) on the NOIZEUS corpus demonstrate that the AKF constructed with MHANet-LPC-PS driven speech and noise LPC parameters produced enhanced speech with higher quality and intelligibility than competing methods.

scholarly journals On Training Targets for Supervised LPC Estimation to Augmented Kalman Filter-based Speech Enhancement

2021 ◽  
Author(s):  
Sujan Kumar Roy ◽  
Aaron Nicolson ◽  
Kuldip K. Paliwal
Keyword(s):  
Power Spectrum ◽  
Speech Enhancement ◽  
Linear Prediction ◽  
Real Life ◽  
Large Data ◽  
Convolutional Network ◽  
Data Set ◽  
Listening Tests ◽  
Objective Quality ◽  
Prediction Coefficient

The performance of speech coding, speech recognition, and speech enhancement largely depends upon the accuracy of the linear prediction coefficient (LPC) of clean speech and noise in practice. Formulation of speech and noise LPC estimation as a supervised learning problem has shown considerable promise. In its simplest form, a supervised technique, typically a deep neural network (DNN) is trained to learn a mapping from noisy speech features to clean speech and noise LPCs. Training targets for DNN to clean speech and noise LPC estimation fall into four categories: line spectrum frequency (LSF), LPC power spectrum (LPC-PS), power spectrum (PS), and magnitude spectrum (MS). The choice of appropriate training target as well as the DNN method can have a significant impact on LPC estimation in practice. Motivated by this, we perform a comprehensive study on the training targets using two state-of-the-art DNN methods--- residual network and temporal convolutional network (ResNet-TCN) and multi-head attention network (MHANet). This study aims to determine which training target as well as DNN method produces more accurate LPCs in practice. We train the ResNet-TCN and MHANet for each training target with a large data set. Experiments on the NOIZEUS corpus demonstrate that the LPC-PS training target with MHANet produces a lower spectral distortion (SD) level in the estimated speech LPCs in real-life noise conditions. We also construct the AKF with the estimated speech and noise LPC parameters from each training target using ResNet-TCN and MHANet. Subjective AB listening tests and seven different objective quality and intelligibility evaluation measures (CSIG, CBAK, COVL, PESQ, STOI, SegSNR, and SI-SDR) on the NOIZEUS corpus demonstrate that the AKF constructed with MHANet-LPC-PS driven speech and noise LPC parameters produced enhanced speech with higher quality and intelligibility than competing methods.

scholarly journals DeepLPC-MHANet: Multi-Head Self-Attention for Augmented Kalman Filter-based Speech Enhancement

2021 ◽  
Author(s):  
Sujan Kumar Roy ◽  
Aaron Nicolson ◽  
Kuldip K. Paliwal
Keyword(s):  
Deep Learning ◽  
Kalman Filter ◽  
Speech Enhancement ◽  
Linear Prediction ◽  
Power Spectra ◽  
Previous Method ◽  
Learning Approaches ◽  
Convolutional Network ◽  
Listening Tests ◽  
Prediction Coefficient

Current augmented Kalman filter (AKF)-based speech enhancement algorithms utilise a temporal convolutional network (TCN) to estimate the clean speech and noise linear prediction coefficient (LPC). However, the multi-head attention network (MHANet) has demonstrated the ability to more efficiently model the long-term dependencies of noisy speech than TCNs. Motivated by this, we investigate the MHANet for LPC estimation. We aim to produce clean speech and noise LPC parameters with the least bias to date. With this, we also aim to produce higher quality and more intelligible enhanced speech than any current KF or AKF-based SEA. Here, we investigate MHANet within the DeepLPC framework. DeepLPC is a deep learning framework for jointly estimating the clean speech and noise LPC power spectra. DeepLPC is selected as it exhibits significantly less bias than other frameworks, by avoiding the use of whitening filters and post-processing. DeepLPC-MHANet is evaluated on the NOIZEUS corpus using subjective AB listening tests, as well as seven different objective measures (CSIG, CBAK, COVL, PESQ, STOI, SegSNR, and SI-SDR). DeepLPC-MHANet is compared to five existing deep learning-based methods. Compared to other deep learning approaches, DeepLPC-MHANet produced clean speech LPC estimates with the least amount of bias. DeepLPC-MHANet-AKF also produced higher objective scores than any of the competing methods (with an improvement of 0.17 for CSIG, 0.15 for CBAK, 0.19 for COVL, 0.24 for PESQ, 3.70\% for STOI, 1.03 dB for SegSNR, and 1.04 dB for SI-SDR over the next best method). The enhanced speech produced by DeepLPC-MHANet-AKF was also the most preferred amongst ten listeners. By producing LPC estimates with the least amount of bias to date, DeepLPC-MHANet enables the AKF to produce enhanced speech at a higher quality and intelligibility than any previous method.

scholarly journals DeepLPC-MHANet: Multi-Head Self-Attention for Augmented Kalman Filter-based Speech Enhancement

2021 ◽  
Author(s):  
Sujan Kumar Roy ◽  
Aaron Nicolson ◽  
Kuldip K. Paliwal
Keyword(s):  
Deep Learning ◽  
Kalman Filter ◽  
Speech Enhancement ◽  
Linear Prediction ◽  
Power Spectra ◽  
Previous Method ◽  
Learning Approaches ◽  
Convolutional Network ◽  
Listening Tests ◽  
Prediction Coefficient

Current augmented Kalman filter (AKF)-based speech enhancement algorithms utilise a temporal convolutional network (TCN) to estimate the clean speech and noise linear prediction coefficient (LPC). However, the multi-head attention network (MHANet) has demonstrated the ability to more efficiently model the long-term dependencies of noisy speech than TCNs. Motivated by this, we investigate the MHANet for LPC estimation. We aim to produce clean speech and noise LPC parameters with the least bias to date. With this, we also aim to produce higher quality and more intelligible enhanced speech than any current KF or AKF-based SEA. Here, we investigate MHANet within the DeepLPC framework. DeepLPC is a deep learning framework for jointly estimating the clean speech and noise LPC power spectra. DeepLPC is selected as it exhibits significantly less bias than other frameworks, by avoiding the use of whitening filters and post-processing. DeepLPC-MHANet is evaluated on the NOIZEUS corpus using subjective AB listening tests, as well as seven different objective measures (CSIG, CBAK, COVL, PESQ, STOI, SegSNR, and SI-SDR). DeepLPC-MHANet is compared to five existing deep learning-based methods. Compared to other deep learning approaches, DeepLPC-MHANet produced clean speech LPC estimates with the least amount of bias. DeepLPC-MHANet-AKF also produced higher objective scores than any of the competing methods (with an improvement of 0.17 for CSIG, 0.15 for CBAK, 0.19 for COVL, 0.24 for PESQ, 3.70\% for STOI, 1.03 dB for SegSNR, and 1.04 dB for SI-SDR over the next best method). The enhanced speech produced by DeepLPC-MHANet-AKF was also the most preferred amongst ten listeners. By producing LPC estimates with the least amount of bias to date, DeepLPC-MHANet enables the AKF to produce enhanced speech at a higher quality and intelligibility than any previous method.

scholarly journals Robustness and Sensitivity Tuning of the Kalman Filter for Speech Enhancement

Signals ◽  
2021 ◽  
Vol 2 (3) ◽  
pp. 434-455
Author(s):  
Sujan Kumar Roy ◽  
Kuldip K. Paliwal
Keyword(s):  
Kalman Filter ◽  
Speech Enhancement ◽  
Linear Prediction ◽  
Real Life ◽  
Model Parameters ◽  
Noise Variance ◽  
Noisy Speech ◽  
Kalman Gain ◽  
Whitening Filter ◽  
Prediction Coefficient

Inaccurate estimates of the linear prediction coefficient (LPC) and noise variance introduce bias in Kalman filter (KF) gain and degrade speech enhancement performance. The existing methods propose a tuning of the biased Kalman gain, particularly in stationary noise conditions. This paper introduces a tuning of the KF gain for speech enhancement in real-life noise conditions. First, we estimate noise from each noisy speech frame using a speech presence probability (SPP) method to compute the noise variance. Then, we construct a whitening filter (with its coefficients computed from the estimated noise) to pre-whiten each noisy speech frame prior to computing the speech LPC parameters. We then construct the KF with the estimated parameters, where the robustness metric offsets the bias in KF gain during speech absence of noisy speech to that of the sensitivity metric during speech presence to achieve better noise reduction. The noise variance and the speech model parameters are adopted as a speech activity detector. The reduced-biased Kalman gain enables the KF to minimize the noise effect significantly, yielding the enhanced speech. Objective and subjective scores on the NOIZEUS corpus demonstrate that the enhanced speech produced by the proposed method exhibits higher quality and intelligibility than some benchmark methods.

scholarly journals DeepLPC: A Deep Learning Approach to Augmented Kalman Filter-Based Single-Channel Speech Enhancement

2021 ◽  
Author(s):  
Sujan Kumar Roy ◽  
Aaron Nicolson ◽  
Kuldip K. Paliwal
Keyword(s):  
Deep Learning ◽  
Kalman Filter ◽  
Linear Prediction ◽  
Single Channel ◽  
Power Spectra ◽  
Learning Approaches ◽  
Learning Framework ◽  
Listening Tests ◽  
Whitening Filter ◽  
Prediction Coefficient

Current deep learning approaches to linear prediction coefficient (LPC) estimation for the augmented Kalman filter (AKF) produce bias estimates, due to the use of a whitening filter. This severely degrades the perceived quality and intelligibility of enhanced speech produced by the AKF. In this paper, we propose a deep learning framework that produces clean speech and noise LPC estimates with significantly less bias than previous methods, by avoiding the use of a whitening filter. The proposed framework, called DeepLPC, jointly estimates the clean speech and noise LPC power spectra. The estimated clean speech and noise LPC power spectra are passed through the inverse Fourier transform to form autocorrelation matrices, which are then solved by the Levinson-Durbin recursion to form the LPCs and prediction error variances of the speech and noise for the AKF. The performance of DeepLPC is evaluated on the NOIZEUS and DEMAND Voice Bank datasets using subjective AB listening tests, as well as seven different objective measures (CSIG, CBAK, COVL, PESQ, STOI, SegSNR, and SI-SDR). DeepLPC is compared to six existing deep learning-based methods. Compared to other deep learning approaches to clean speech LPC estimation, DeepLPC produces a lower spectral distortion (SD) level than existing methods, confirming that it exhibits less bias. DeepLPC also produced higher objective scores than any of the competing methods (with an improvement of 0.11 for CSIG, 0.15 for CBAK, 0.14 for COVL, 0.13 for PESQ, 2.66\% for STOI, 1.11 dB for SegSNR, and 1.05 dB for SI-SDR, over the next best method). The enhanced speech produced by DeepLPC was also the most preferred by listeners. By producing less biased clean speech and noise LPC estimates, DeepLPC enables the AKF to produce enhanced speech at a higher quality and intelligibility.

scholarly journals Blind Queries Applied to JSON Document Stores

Information ◽  
2019 ◽  
Vol 10 (10) ◽  
pp. 291 ◽  
Author(s):  
Stefania Marrara ◽  
Mauro Pelucchi ◽  
Giuseppe Psaila
Keyword(s):  
Real Life ◽  
Open Data ◽  
Large Data ◽  
General Information ◽  
Data Sets ◽  
Heterogeneous Structure ◽  
Data Set ◽  
Document Database ◽  
Job Vacancy ◽  
Document Stores

Social Media, Web Portals and, in general, information systems offer their own Application Programming Interfaces (APIs), used to provide large data sets concerning every aspect of day-by-day life. APIs usually provide data sets as collections of JSON documents. The heterogeneous structure of JSON documents returned by different APIs constitutes a barrier to effectively query and analyze these data sets. The adoption of NoSQL document stores, such as MongoDB, is useful for gathering these data sets, but does not solve the problem of querying the final heterogeneous repository. The aim of this paper is to provide analysts with a tool, named HammerJDB, that allows for blind querying collections of JSON documents within a NoSQL document database. The idea below is that users may know the application domain but it may be that they are not aware of the real structures of the documents stored in the database—the tool for blind querying tries to bridge the gap, by adopting a query rewriting mechanism. This paper is an evolution of a technique for blind querying Open Data portals and of its implementation within the Hammer framework, presented in some previous work. In this paper, we evolve that approach in order to query a NoSQL document database by evolving the Hammer framework into the HammerJDB framework, which is able to work on MongoDB databases. The effectiveness of the new approach is evaluated on a data set (derived from a real-life one), containing job-vacancy ads collected from European job portals.

scholarly journals Robustness and Sensitivity Tuning of the Kalman Filter for Speech Enhancement

2021 ◽  
Author(s):  
Sujan Kumar Roy ◽  
Kuldip K. Paliwal
Keyword(s):  
Kalman Filter ◽  
Speech Enhancement ◽  
Linear Prediction ◽  
Real Life ◽  
Model Parameters ◽  
Noise Variance ◽  
Noise Condition ◽  
Noisy Speech ◽  
Kalman Gain ◽  
Whitening Filter

The inaccurate estimates of linear prediction coefficient (LPC) and noise variance introduce bias in Kalman filter (KF) gain and degrades speech enhancement performance. The existing methods proposed a tuning of the biased Kalman gain particularly in stationary noise condition. This paper introduces a tuning of the KF gain for speech enhancement in real-life noise conditions. First, we estimate noise from each noisy speech frame using a speech presence probability (SPP) method to compute the noise variance. Then construct a whitening filter (with its coefficients computed from the estimated noise) and employed to the noisy speech, yielding a pre-whitened speech, from where the speech LPC parameters are computed. Then construct KF with the estimated parameters, where the robustness metric offsets the bias in Kalman gain during speech absence to that of the sensitivity metric during speech presence to achieve better noise reduction. Where the noise variance and the speech model parameters are adopted as a speech activity detector. The reduced-biased Kalman gain enables the KF to minimize the noise effect significantly, yielding the enhanced speech. Objective and subjective scores on NOIZEUS corpus demonstrates that the enhanced speech produced by the proposed method exhibits higher quality and intelligibility than some benchmark methods.

scholarly journals UNSUPERVISED LEARNING BASED DISTRIBUTED DETECTION OF GLOBAL ANOMALIES

2010 ◽  
Vol 09 (06) ◽  
pp. 935-957 ◽  
Author(s):  
JUNLIN ZHOU ◽  
ALEKSANDAR LAZAREVIC ◽  
KUO-WEI HSU ◽  
JAIDEEP SRIVASTAVA ◽  
YAN FU ◽  
...  
Keyword(s):  
Anomaly Detection ◽  
Real Life ◽  
Large Data ◽  
Detection Algorithm ◽  
Detection Methods ◽  
Communication Overhead ◽  
Distributed Data ◽  
Local Models ◽  
Data Set ◽  
Global Anomaly

Anomaly detection has recently become an important problem in many industrial and financial applications. Very often, the databases from which anomalies have to be found are located at multiple local sites and cannot be merged due to privacy reasons or communication overhead. In this paper, a novel general framework for distributed anomaly detection is proposed. The proposed method consists of three steps: (i) building local models for distributed data sources with unsupervised anomaly detection methods and computing quality measure of local models; (ii) transforming local unsupervised local models into sharing models; and (iii) reusing sharing models for new data and combining their results by considering both quality and diversity of them to detect anomalies in a global view. In experiments performed on synthetic and real-life large data set, the proposed distributed anomaly detection method achieved prediction performance comparable or even slightly better than the global anomaly detection algorithm applied on the data set obtained when all distributed data set were merged.

Research on Optimized Clustering Analysis Algorithm

2013 ◽  
Vol 811 ◽  
pp. 407-412
Author(s):  
Lin Tong ◽  
Di Wu
Keyword(s):  
Clustering Analysis ◽  
Real Life ◽  
Large Data ◽  
Data Set ◽  
Analysis Algorithm ◽  
The Poor ◽  
Large Data Set

In this paper an improved clustering analysis algorithm is proposed on the basis of randomly selected data in the CURE algorithm and the cluster centers setting in the K-NN algorithm. The combination of the two algorithms conquers the poor clustering accuracy in the CURE algorithm and the clustering deficiency on large data set in the K-NN algorithm. This paper first introduces the concept of clustering analysis and its real-life applications, then proceeds to describe the method of clustering analysis, and then introduces optimized CURE_KNN algorithm described in pseudo-code, at last the advantages of the algorithm are summarized.

Improving discretization based pattern discovery for multivariate time series by additional preprocessing

2021 ◽  
Vol 25 (5) ◽  
pp. 1051-1072
Author(s):  
Fabian Kai-Dietrich Noering ◽  
Konstantin Jonas ◽  
Frank Klawonn
Keyword(s):  
Time Series ◽  
Time Series Data ◽  
Multivariate Time Series ◽  
Pattern Discovery ◽  
Computing Time ◽  
Real Life ◽  
Large Data ◽  
Series Data ◽  
Data Set ◽  
Recurrent Patterns

In technical systems the analysis of similar load situations is a promising technique to gain information about the system’s state, its health or wearing. Very often, load situations are challenging to be defined by hand. Hence, these situations need to be discovered as recurrent patterns within multivariate time series data of the system under consideration. Unsupervised algorithms for finding such recurrent patterns in multivariate time series must be able to cope with very large data sets because the system might be observed over a very long time. In our previous work we identified discretization-based approaches to be very interesting for variable length pattern discovery because of their low computing time due to the simplification (symbolization) of the time series. In this paper we propose additional preprocessing steps for symbolic representation of time series aiming for enhanced multivariate pattern discovery. Beyond that we show the performance (quality and computing time) of our algorithms in a synthetic test data set as well as in a real life example with 100 millions of time points. We also test our approach with increasing dimensionality of the time series.

Sign in / Sign up

Export Citation Format

Share Document