Nam-hoon Kim

Learning feature mapping using deep neural network bottleneck features for distant large vocabulary speech recognition

Automatic speech recognition from distant microphones is a difficult task because recordings are affected by reverberation and background noise. First, the application of the deep neural network (DNN)/hidden Markov model (HMM) hybrid acoustic models for distant speech recognition task using AMI meeting corpus is investigated. This paper then proposes a feature transformation for removing reverberation and background noise artefacts from bottleneck features using DNN trained to learn the mapping between distant-talking speech features and close-talking speech bottleneck features. Experimental results on AMI meeting corpus reveal that the mismatch between close-talking and distant-talking conditions is largely reduced, with about 16% relative improvement over conventional bottleneck system (trained on close-talking speech). If the feature mapping is applied to close-talking speech, a minor degradation of 4% relative is observed.

RightTopologizer: an efficient schematic generator for multi-level optimization

This paper presents an efficient algorithm for the automatic generation of schematic diagrams for digital systems described not only at the gate-level but also at the RTL. Schematics are generated in the sequence of pre-processing, topological placement, geometrical placement, and channel routing. The experimental results show that RightTopologizer produces more readable and compact schematics than a well-known commercial tool in a real time response. The high readability and compactness of the generated schematics helps the designers make early design decisions by providing a good grasp of signal flows which represent HDL-based designs. In addition, this system can be used for a multi-level optimization framework in a variety of applications due to using a common database with our schematic capture system.

Accent adaptation using Subspace Gaussian Mixture Models

This paper investigates employment of Subspace Gaussian Mixture Models (SGMMs) for acoustic model adaptation towards different accents for English speech recognition. The SGMMs comprise globally-shared and state-specific parameters which can efficiently be employed for various kinds of acoustic parameter tying. Research results indicate that well-defined sharing of acoustic model parameters in SGMMs can significantly outperform adapted systems based on conventional HMM/GMMs. Furthermore, SGMMs rapidly achieve target acoustic models with small amounts of data. Experiments performed with US and UK English versions of the Wall Street Journal (WSJ) corpora indicate that SGMMs lead to approximately 20% and 8% relative improvements with respect to speaker-independent and speaker-adapted acoustic models respectively over conventional HMM/GMMs. Finally, we demonstrate that SGMMs adapted only with 1.5 hours can reach performance of HMM/GMMs trained with 18 hours.

The Silurian-Devonian magmatism recorded in detrital zircons from the Andong area, northeastern Yeongnam Massif, Korea

We report in situ U-Pb ages and Hf isotopic compositions of detrital zircons in three metasedimentary gneisses from the Andong area, northeastern Yeongnam Massif, Korea. Zircons from two xenoliths entrained in ca. 250 Ma tonalitic-granodioritic metagranitoids respectively show predominant Silurian clusters slightly overlapping each other within uncertainties (428 ± 7 and 419 ± 10 Ma), together with subordinate Archean-Proterozoic peaks. A younger Devonian population (372 ± 5 Ma) is yielded by the youngest and most dominant age component in a biotite gneiss of the gneiss-marbleamphibolite complex occurring to the southeast of the metagranitoids. The magmatic origin of these zircons is corroborated by their oscillatory cathodoluminescence zoning and relatively high Th/U ratios (0.6 ± 0.4). Therefore, the depositional timing of the protolith of the gneisses is constrained to be Late Paleozoic. Scattered metamorphic rim dates of the detrital zircons (ca. 260–220 Ma) reflect post-depositional thermal overprints resulting from recurrent arc-related magmatism. Highly variable initial εHf values of the Silurian zircons (+6 to −18) indicate contributions from both primitive melts and pre-existing crustal materials, which are typical for continental arc magmas. The Silurian-Devonian magmatism recorded in the detrital zircons is possibly linked with the Pb loss event(s) inferred from the zircon data of the Paleoproterozoic basement rocks of northeastern Yeongnam Massif and the overlying Cambrian sediments.

Integration of sporadic noise model in POMDP-based voice activity detection

Partially observable Markov decision process (POMDP) has recently been applied to a voice activity detector (VAD), which makes it possible to incorporate knowledge about the recording environments in the decision process in order to achieve a more stable performance in noisy situations. In this paper, the model has been further explored to utilize prior knowledge about possible intermittent noise signals such as breath or click sounds, in addition to the stationary background noise types. The experimental result shows that application of sporadic noise models in a POMDP-based VAD reduces the equal error rate of the voice activity decision by about 7% relatively.

How to extend battery life in location alarm

Location Alarm application notifying user of an alarm at the locations where user registered events causes serious battery drain because the current location should be continuously monitored with turning on the GPS or WPS. This paper proposes a scheme to extend battery life that minimizes the access to the GPS or WPS considering the users moving state recognized by various sensors in mobile device.