T. Kishore Kumar

Speech enhancement using kernel adaptive filtering method

In this paper, we investigate the enhancement of speech by applying kernel adaptive filter. Noise removal is very important in many applications like telephone conversation, speech recognition, etc. Kernel methods have shown good results for other applications like handwriting recognition, inverse distance weightings, etc. To improve the speech quality and intelligibility, we can process the signals in new domain like Reproducing Kernel Hilbert Space (RKHS) unlike time and frequency domains. We have used the noisy speech corpus (NOIZEUS) for the experiments. The experimental results shown the noise removal in RKHS has good improvement in the Signal to Noise Ratio (SNR) values as compared the traditional methods.

Speech Recognition Using Neural Networks

Speech processing has been an active area for several decades with a wide variety of applications ranging from communications to automatic reading machines. There are many speech recognition techniques, which are based on statistical techniques as well as neural networks. The present work investigates the feasibility of two approaches for solving the problem using Neural Networks.

Particle Swarm Optimization Based Tuning of Unscented Kalman Filter for Bearings Only Tracking

Kalman filter is a well known adaptive filtering Algorithm, widely used for target tracking applications. When the system model and measurements are non linear, variation of Kalman filter like extended Kalman filter (EKF) and Unscented Kalman filters (UKF) are used. For obtaining reliable estimate of the target state, filter has to be tuned before the operation (off line). Tuning an UKF is the process of estimation of the noise covariance matrices from process data. In practical applications, due to unavailable measurements of the process noise and high dimensionality of the problem tuning of the filter is left for engineering intuition. In this paper, tuning of the UKF is investigated using Particle Swarm Optimization (PSO). The simulation results show the superiority of the PSO tuned UKF over the conventional tuned UKF.

A Novel Approach for Clutter Reduction in Through-the-wall UWB Radar Imaging

Abstract Through-the-wall imaging (TWI) is an emerging area of research and development which is very much useful in urban warfare situations. Ultra-wide band range radar is the best suitable for this application where this range of signals has the capability of penetrating through the materials but still can provide better resolution. The incorporation of signal processing techniques on the raw data will give better representation of the scanned scene. The paper uses impulse radar for TWI, and back projection algorithm is used to retrieve the B-scan signal. A constant gain of factor 100 is used to increase the echo strength. Singular value decomposition is used to reduce the clutter from the B-scan signal and the peak signal to noise ratio is observed to be −3.79 dB.

Speech enhancement by Bayesian estimation of clean speech modeled as super Gaussian given a priori knowledge of phase

Maximum Likelihood (ML), Maximum a posterior (MAP) and Minimum mean square error (MMSE) estimators are proposed for speech enhancement.Noise spectral amplitude priors assumed as Laplace, Gaussian probability density functions (pdf).Speech spectral amplitude priors as Nakagami, Gamma distributed.A joint MMSE estimate of the Complex speech coefficients given Uncertain Phase information is derived. In this paper, STFT based speech enhancement algorithms based on estimation of short time spectral amplitudes are proposed. These algorithms use Maximum Likelihood (ML), Maximum a posterior (MAP) and Minimum mean square error (MMSE) estimators which respectively uses Laplace, Gaussian probability density functions (pdf) as noise spectral amplitude priors and Nakagami, Gamma distributions as speech spectral amplitude priors. The method uses a joint MMSE estimate of the clean speech amplitude and clean speech phase for a given uncertainty phase information for improved single channel speech enhancement. In the most of the speech enhancement algorithms, we only concentrate on the frequency domain amplitude of speech, but not on the phase of noisy speech since it may cause undesired artifacts. In this paper, a recent phase reconstruction algorithm is used to estimate the phase of clean speech. The reconstructed phase is treated as an uncertain prior knowledge when deriving a joint MMSE estimate of the Complex speech coefficients given Uncertain Phase (CUP) information. The proposed MMSE optimal CUP estimator reduces undesired artifacts and also gives satisfactory values between the phase of noisy signal and the estimate of prior phase. We evaluate all the above estimators using speech signals uttered by 10 male speakers and 10 female speakers are taken from TIMIT database. The proposed method outperforms other benchmark algorithms in terms of segmental signal to noise ratio (SSNR), short-time objective intelligibility (STOI) and perceptual evaluation of speech quality (PESQ).

Speech enhancement using sub-band wiener filter with pitch synchronous analysis

This paper introduces a novel speech enhancement system based on sub-band wiener filter with pitch synchronous analysis. The perceptual filter bank provides a good auditory representation, good perceptual quality of speech. Sub-band wiener filter based Pitch synchronous analysis reduces drawbacks of fixed window shift. To increase the inter frame similarities the shift of analysis window is based on pitch period. The pitch is extracted by using clipping level method. Further, to reduce noise each sub-band is wiener filtered using a priori SNR with adaptive parameter. Wavelets transform based wiener filter approach works well than DCT based approach. Objective (SNR, segSNR, LLR, LSD, IS,WSS) experiments prove that the new speech enhancement system is capable of significant noise reduction.

UWB Impulse Radar for Through-The-Wall Imaging

Ultra-Wideband (UWB) is the promising technology for localization of the objects behind the walls. Recent terrorist activities and law-enforcement situations underscore the need for effective through-wall detection. UWB radar signals has extremely large frequency spectrum and since low frequencies has more penetration capabilities through dielectric materials it is best suitable for Through-the-Wall Radar Imaging (TWRI). Signal processing in TWRI has a greater impact in getting the information of the scanned area. This paper uses impulse signals in TWRI, examines the factors impacting in TWRI and obtains the two dimensional information of the scanned scene. Electromagnetic simulation software is used to generate the room like structure, and to obtain the raw radar data.

Speech enhancement using posterior regularized NMF with bases update

Abstract In this paper, a combination of statistical model-based approach and Non-negative Matrix Factorization (NMF)-based approach with on-line update of speech and noise bases for speech enhancement is proposed. Template-based approaches are more robust and perform better than non-stationary noises compared to statistical model-based approaches but are dependent on a priori information. Combining the approaches avoids the drawbacks of both. To improve the performance further, speech and noise bases are adapted simultaneously in NMF approach with the help of the estimated speech presence probability (SPP). The proposed method outperforms other benchmark algorithms in terms of perceptual evaluation of speech quality (PESQ) and source-to-distortion ratio (SDR) in stationary and non-stationary noise environment conditions with matched and mismatched noise basis.

An informed separation algorithm based on sound field mapping for speech recognition systems

This paper considers the problem of separation of speech sources from signals captured by a microphone array, and its impact on speech recognition systems. The proposed method improves the novel separation procedure based on processing of sound field maps in the ray space by incorporating the estimates of signal spectral envelopes into the design of the separation filter. In particular, we resort to a two-stage algorithm in which the assumptions regarding the signal spectral densities made in the first stage are replaced by their parametric estimates in the second stage. The performance gain is evaluated in real experiments and the impact on speech recognition accuracy is examined using several commercial cloud-based speech recognition APIs.

Efficient pitch detection algorithms for pitched musical instrument sounds: A comparative performance evaluation

Pitch detection of an audio signal is an interesting research topic in the field of speech signal processing. Pitch is one of the most important perceptual features, as it conveys much information about the audio signal. It is closely related to the physical feature of fundamental frequency f0. For musical instrument sounds, the f0 and the measured pitch can be considered equivalent. In this paper four pitch detection algorithms have been proposed for pitched musical instrument sounds. The goal of this paper is to investigate how these algorithms should be adapted to pitched musical instrument sounds analysis and to provide a comparative performance evaluation of the most representative state-of-the-art approaches. This study is carried out on a large database of pitched musical instrument sounds, comprising four types of pitched musical instruments violin, trumpet, guitar and flute. The algorithmic performance is assessed according to the ability to estimate pitch contour accurately.