Ching-Chung Li

Speech signal modification to increase intelligibility in noisy environments

The role of transient speech components on speech intelligibility was investigated. Speech was decomposed into two components--quasi-steady-state (QSS) and transient--using a set of time-varying filters whose center frequencies and bandwidths were controlled to identify the strongest formant components in speech. The relative energy and intelligibility of the QSS and transient components were compared to original speech. Most of the speech energy was in the QSS component, but this component had low intelligibility. The transient component had much lower energy but was almost as intelligible as the original speech, suggesting that the transient component included speech elements important to speech perception. A modified version of speech was produced by amplifying the transient component and recombining it with the original speech. The intelligibility of the modified speech in background noise was compared to that of the original speech, using a psychoacoustic procedure based on the modified rhyme protocol. Word recognition rates for the modified speech were significantly higher at low signal-to-noise ratios (SNRs), with minimal effect on intelligibility at higher SNRs. These results suggest that amplification of transient information may improve the intelligibility of speech in noise and that this improvement is more effective in severe noise conditions.

Efficient computation of the discrete pseudo-Wigner distribution

A description is given of a novel algorithm, the fast Fourier transform in part (FFTP), for the computation of the discrete pseudo-Wigner distribution (DPWD). The FFTP computes the cosine and sine parts of the discrete Fourier transform (DFT) separately by employing real inverse sinusoidal twiddle factors. Unlike the conventional methods which directly utilize the complex DFT, the FFTP yields real output since the DPWD is always real. In addition, the new method reduces the computational cost by making full use of symmetries and removing redundancies in the FFTP computation. The authors also describe a simple algorithm for computing the discrete Hilbert transform (DHT) to produce the nonaliased DPWD. A pipeline structure for real-time and a bulk processing technique for offline implementations of the method are presented. >

Wavelet denoising via sparse representation

Wavelet threshold denoising is a powerful method for suppressing noise in signals and images. However, this method often uses a coordinate-wise processing scheme, which ignores the structural properties in the wavelet coefficients. We propose a new wavelet denoising method using sparse representation which is a powerful mathematical tool recently developed. Instead of thresholding wavelet coefficients individually, we minimize the number of non-zero coefficients under certain conditions. The denoised signal is reconstructed by solving an optimization problem. It is shown that the solution to the optimization problem can be obtained uniquely and the estimates of the denoised wavelet coefficients are unbiased, i.e., the statistical means of the estimates are equal to the noise-free wavelet coefficients. It is also shown that at least a local optimal solution to the denoising problem can be found. Our experiments on test data indicate that this new denoising method is effective and efficient for a wide variety of signals including those with low signal-to-noise ratios.

New signal decomposition method based speech enhancement

The auditory system, like the visual system, may be sensitive to abrupt stimulus changes, and the transient component in speech may be particularly critical to speech perception. If this component can be identified and selectively amplified, improved speech perception in background noise may be possible. This paper describes an algorithm to decompose speech into tonal, transient, and residual components. The modified discrete cosine transform (MDCT) was used to capture the tonal component and the wavelet transform was used to capture transient features. A hidden Markov chain (HMC) model and a hidden Markov tree (HMT) model were applied to capture statistical dependencies between the MDCT coefficients and between the wavelet coefficients, respectively. The transient component identified by the wavelet transform was selectively amplified and recombined with the original speech to generate modified speech, with energy adjusted to equal the energy of the original speech. The intelligibility of the original and modified speech was evaluated in eleven human subjects using the modified rhyme protocol. Word recognition rate results show that the modified speech can improve speech intelligibility at low SNR levels (8% at -15dB, 14% at -20dB, and 18% at -25dB) and has minimal effect on intelligibility at higher SNR levels.

A wavelet based neural network for prediction of ICP signal

We present a wavelet-based neural network for multi-step prediction of the intracranial pressure (ICP) signal. A multiresolution dynamic predictor (MDP) is proposed, which utilizes the discrete wavelet transform computing wavelet coefficients from coarse scale to fine scale and recurrent neural networks (RNNs) forming dynamic nonlinear models for prediction. It has the ability to predict the ICP in both long-term with coarse resolution and short-term with fine resolution. Computational results up to three scale levels have demonstrated the effectiveness of the MDP for multi-step prediction as compared with the the raw data.

A secret image sharing method using integer wavelet transform

A new image sharing method, based on the reversible integer-to-integer (ITI) wavelet transform and Shamirs threshold scheme is presented, that provides highly compact shadows for real-time progressive transmission. This method, working in the wavelet domain, processes the transform coefficients in each subband, divides each of the resulting combination coefficients into shadows, and allows recovery of the complete secret image by using any or more shadows . We take advantages of properties of the wavelet transform multiresolution representation, such as coefficient magnitude decay and excellent energy compaction, to design combination procedures for the transform coefficients and processing sequences in wavelet subbands such that small shadows for real-time progressive transmission are obtained. Experimental results demonstrate that the proposed method yields small shadow images and has the capabilities of real-time progressive transmission and perfect reconstruction of secret images.

Speech Enhancement Using Transient Speech Components

This paper describes an algorithm to decompose speech into tonal, transient, and residual components. The algorithm uses an MDCT-based hidden Markov chain model to isolate the tonal component and a wavelet-based hidden Markov tree model to isolate the transient component. We suggest that the auditory system, like the visual system, is probably sensitive to abrupt stimulus changes and that the transient component in speech may be particularly critical to speech perception. To test this suggestion, the transient component isolated by our algorithm was selectively amplified and recombined with the original speech to generate enhanced speech, with energy adjusted to be equal to the energy of the original speech. The intelligibility of the original and enhanced speech was evaluated in eleven human subjects by the modified rhyme protocol. The word recognition rates show that the enhanced speech can provide substantial improvement in speech intelligibility at low SNR levels (8% at -15 dB, 14% at -20 dB, and 18% at -25 dB)

A comparative study of two biorthogonal wavelet transforms in time series prediction

We present comparative results of time-series prediction preprocessed with two different wavelet transforms: (1) the compactly supported biorthogonal wavelet transform developed by Cohen, Daubechies, and Feauveau (1992), and (2) the semi-orthogonal wavelet transform, a class of biorthogonal wavelet transform, constructed by Cai and Wang (1996). Both theoretical and computational results of the two wavelet transforms are discussed. The major difference between the two wavelet transforms is the computational procedure. So far, only the semi-orthogonal wavelet transform of Cai and Wang can compute wavelet coefficients from a coarse scale level to a fine scale level, which makes the computation more flexible and cost effective. However, the compactly supported biorthogonal wavelet transform of Cohen et al. Has better decorrelation property. Thus, we found that the semi-orthogonal wavelet transform of Cai and Wang provides a faster computation process while the compactly supported biorthogonal wavelet transform provides better predicted wavelet coefficients in our experimental results. Based on the wavelet coefficients computed from signals, nonlinear prediction models utilizing recurrent neural networks are applied to predict wavelet coefficients at each scale level, Thus, the predicted signal is obtained from the reconstruction of predicted wavelet coefficients, In our experiments, the multi-step prediction using wavelet transforms gives much superior results than those obtained without using wavelet transforms. We applied our method to predict specific time series, intracranial pressure, acquired from head-trauma patients in the neuro intensive care unit at the University of Pittsburgh Medical Center.

Elimination of cross-components of the discrete pseudo Wigner distribution via image processing

The authors introduce an autocomponent selection (ACS) algorithm to remove the cross-components produced by the discrete pseudo-Wigner distribution (DPWD). The ACS treats the DPWD as an image with polarity. This image is processed with an averaging filter to eliminate negative values. The result is then compared with a preprocessed DPWD to classify each image pixel. This approach yields a subset of the DPWD free of redundancies. Unlike traditional smoothing technique, this algorithm does not reduce the time-frequency resolution.<<ETX>>

Automatic Detection of Region of Interest Based on Object Tracking in Neurosurgical Video

Automatic detection of region of interest (ROIs) in a complex image or video, such as an angiogram or endoscopic neurosurgery video, is a critical task in many medical image and video processing applications. In this paper, we present a new method that addresses several challenges in automatic detection of ROI of neurosurgical video for ROI coding which is used for neurophysiological intraoperative monitoring (IOM) system. This method is based on an object tracking technique with multivariate density estimation theory, combined with the shape information of the object. By defining the ROIs for neurosurgical video, this method produces a smooth and convex emphasis region within which surgical procedures are performed. A large bandwidth budget is assigned within the ROI to archive high-fidelity Internet transmission. Outside the ROI, a small bandwidth budget is allocated to efficiently utilize the bandwidth resource . We believe this method also can be used to image-guidance surgery (IGS) systems to track the positions of surgical instruments in the physical space occupied by the patient after some improvement