Kyung-tae Kim

Automatic detection of auditory salience with optimized linear filters derived from human annotation

Auditory salience describes how much a particular auditory event attracts human attention. Previous attempts at automatic detection of salient audio events have been hampered by the challenge of defining ground truth. In this paper ground truth for auditory salience is built up from annotations by human subjects of a large corpus of meeting room recordings. Following statistical purification of the data, an optimal auditory salience filter with linear discrimination is derived from the purified data. An automatic auditory salience detector based on optimal filtering of the Bark-frequency loudness performs with 32% equal error rate. Expanding the feature vector to include other common feature sets does not improve performance. Consistent with intuition, the optimal filter looks like an onset detector in the time domain.

RECOG: A Sensing-Based Cognitive Radio System with Real-Time Application Support

While conventional cognitive radio (CR) system is striving at providing best possible protections for the usage of primary users (PU), little attention has been given to ensure the quality of service (QoS) of applications of secondary users (SU). When loading real-time applications over such a CR system, we have found that existing spectrum sensing schemes create a major hurdle for real-time traffic delivery of SU. For example, energy detection based sensing, a widely used technique, requires possibly more than 100 ms to detect a PU with weak signals. The delay is intolerable for real-time applications with stringent QoS requirements, such as voice over internet protocol (VoIP) or live video chat. This delay, along with other delays caused by backup channel searching, channel switching, and possible buffer overflow due to the insertion of sensing periods, makes supporting real-time applications over CR system very difficult if not impossible. In this paper, we present the design and implementation of a sensing-based CR system - RECOG, which is able to support realtime communications among SUs. We first redesign the conventional sensing scheme. Without increasing the complexity or trading off the detection performance, we break down a long sensing period into a series of shorter blocks, turning a disruptive long delay into negligible short delays. To enhance the sensing capability as well as better protect the QoS of SU traffic, we also incorporate an on-demand sensing scheme based on MAC layer information. In addition, to ensure a fast and reliable switching when PU returns, we integrate an efficient backup channel scanning and searching component in our system. Finally, to overcome a potential buffer overflow, we propose a CR-aware QoS manager. Our extensive experimental evaluations validate that RECOG can not only support realtime traffic among SUs with high quality, but also improve protections for PUs.

Perceptual relevance of the temporal envelope to the speech signal in the 4–7kHz band

The perceptual relevance of adopting the temporal envelope to model the frequency band of 4-7 kHz (highband) in wideband speech signal is described in this letter. Based on theoretical work in psychoacoustics, we find out that the temporal envelope can indeed be a perceptual cue for the high-band signal, i.e., a noiseless sound can be obtained if the temporal envelope is roughly preserved. Subjective listening tests verify that transparent quality can be obtained if the model is used for the 4.5-7 kHz band. The proposed model has the benefits of offering flexible scalability and reducing the cost for quantization in coding applications.

A new adaptive turbo equalizer with soft information classification

Linear turbo equalizers with/without channel estimation have been exploited due to their good performance with low complexity compared to a maximuma posteriori (MAP) turbo equalizer. Much work has focused on channel estimate-based minimum mean square error (MMSE) turbo equalizers. However, an MMSE turbo equalizer still requires higher complexity than an adaptive turbo equalizer such as with a normalized least mean square (NLMS) turbo equalizer. Even if adaptive turbo equalizers converge, there is often a performance loss compared to an MMSE turbo equalizer because the adaptive turbo equalizers treat soft decision data as stationary. In order to reduce this loss, we propose a new adaptive turbo equalizer that uses the soft decision data to switch among a set of K different equalizers to approximate the time varying MMSE behavior. Simulations show that the proposed switching-based NLMS turbo equalizer has better bit error rate (BER) performance than a conventional NLMS turbo equalizer by as much as 0.6dB.

A 90nm generation NOR flash multilevel cell (MLC) with 0.44/spl mu/m/sup 2//bit cell size

A 256Mb NOR MLC flash memory with 90nm technology has been successfully developed. Through judicious integration to control the cell dispersion and charge loss/gain with cycling, we confirm a successful MLC operation up to 10K cycling for 0.44 /spl mu/m/sup 2//bit cell size. In this paper, the key features governing multilevel cell (MLC) operation below 90nm technology node is discussed with experimental results.

Speech enhancement beyond minimum mean squared error with perceptual noise shaping.

Residual error signal after speech enhancement through linear filtering can be decomposed into two disjoint portions: speech signal distortion and background noise suppression. Speech is known to follow a super‐Gaussian probabilistic distribution function (PDF) such as Laplacian, while background noise follows Gaussian PDF. Minimum mean squared error estimation requires only second order statistics not only for the noise but also for the speech. Therefore higher‐order dependence of observed speech on the original speech may cause leakage of speech information into the error residual. This talk will formulate an optimization problem minimizing higher‐order statistics (HOS) as well as energy of the signal distortion constrained by a limit on the maximum audibility of the residual noise. Note that due to the non‐stationary nature of speech, we perform the speech enhancement in short overlapping frames. Minimizing HOS of the speech distortion ensures that the speech distortion includes only noise terms, with ...