Yonghyun Baek

Nonlinear Acoustic Echo Cancellation Using a Nonlinear Postprocessor With a Linearly Constrained Affine Projection Algorithm

This brief presents a nonlinear acoustic echo cancellation algorithm comprising a postprocessor that can model the saturation-type nonlinear distortion of a microphone circuit. To support fast adaptation to the acoustic echo environment, the linear filter is adapted using an affine projection algorithm where the reference input vectors subject to linear operation are dynamically selected to update the coefficient vector. We consider the hard-clipper and sigmoid functions for the postprocessor, and their parameters are adaptively updated using the least-mean-square algorithm. The simulation results confirm that the proposed algorithm achieves fast and stable convergence under saturation-type microphone distortion, and it can have better performance than conventional algorithms.

A priori SAP estimator based on the magnitude square coherence for dual-channel microphone system

In this paper, we present a time-frequency (TF)-dependent a priori speech absence probability (SAP) estimator utilizing the magnitude square coherence (MSC) between two microphone signals. It is shown that the normalized SNR can be numerically computed from the MSC by solving a quadratic equation. Based on the fact that the normalized SNR is bounded between 0 and 1, we directly use it for the probability of speech absence in each TF-unit. Since this approach does not require prior statistical knowledge of noise and speech, it is not affected by the performance of the noise PSD estimator. Furthermore, unlike the conventional SNR-based estimator, additional mapping strategy is unnecessary. The algorithm was evaluated using the receiver operating characteristic (ROC) curve and it attained higher correct detection rate at a given false-alarm rate than the conventional algorithms.

Stereo upmix-based binaural auralization for mobile devices

This paper presents a binaural auralization for mobile device, which is based on a stereo-to-multichannel upmix technique. A modified PCA method in addition to the triple-wise nonnegative amplitude panning are utilized. By adding early reflections, not only the listener envelopment is enriched, but also out-of-head localization is achieved. Subjective test results validate the efficiency of the proposed algorithm.

A high performance hearing aid system with fully programmable ultra low power DSP

This paper presents the design of fully programmable digital signal processor for the hearing aid system and optimized implementation of digital hearing aid algorithms. The average performance of the developed processor is measured by performing Fast Fourier Transform (FFT) algorithm. The proposed applications such as variable multiband loudness compensation algorithm with wide dynamic range compression and sub-band based adaptive feedback cancellation algorithm are optimized for real-time processing by exploiting the data and instruction parallelism of the developed processor. The results show that the computational ability and power consumption of the developed processor are suitable to use for the hearing aid system. The optimization achieves that the computational time to perform the full applications on the processor takes only under 40% of the constraint for real-time processing.

Acoustic distance rendering for the application of internet-based mobile virtual reality games

In this paper, we propose a low-complexity acoustic distance rendering (ADR) algorithm useful for the internet-based mobile virtual reality (VR) game. The proposed algorithm approximates the binaural distance variation function (DVF) in [3] which transforms the far-field HRTFs to near-field ones. Near-filed effects due to the distance-and frequency-dependent inter-aural level difference (ILD) and head shadowing are simplified using a distance ratio and a first order IIR filter. Subjective and objective test results show that the spectral distortion (SD) caused by the approximation is under 0.5dB and the proposed algorithm is perceptually equivalent to the DVF-based distance rendering.

Robust Primary-ambient Signal Decomposition Method using Principal Component Analysis with Phase Alignment

The primary and ambient signal decomposition of a stereo sound is a key step to the stereo upmix. The principal component analysis (PCA) is one of the most widely used methods of primary-ambient signal decomposition. However, previous PCA-based decomposition algorithms assume that stereo sound sources are only amplitude-panned without any consideration of phase difference. So it occurs some performance degradation in case of live recorded stereo sound. In this paper, we propose a new PCA-based stereo decomposition algorithm that can consider the phase difference between the channel signals. The proposed algorithm overcomes limitation of conventional signal model using PCA with phase alignment. The phase alignment is realized by using inter-channel phase difference (IPD) which is widely used in parametric stereo coding. Moreover, Enhanced Modified PCA(EMPCA) is combined to solve the problem of conventional PCA caused by Primary to Ambient energy Ratio(PAR) and panning angle dependency. The simulation results are presented to show the improvements of the proposed algorithm.

Real-time binaural noise reduction in diffuse noise field

In this paper, we present a real-time noise reduction algorithm for binaural sound devices operating in diffuse noise fields. The diffuse noise PSD is estimated in a form of the minimum eigenvalue of the (2×2) input covariance matrix. To solve the under-estimation problem at low-frequencies, a compensation scheme is used. Also, to alleviate spectral bias due to smoothing, the estimated noise PSD is scaled by a smoothing factor. The proposed algorithm was implemented in real-time using a 16-bit programmable DSP core. Real-time implementation indicates that the algorithm requires less than 4MHz clock, so that it is suitable for wearable sound devices such as binaural hearing aids and headphones with noise reduction features.