Mingu Lee

A bit reduction algorithm for Spectral Band Replication based on human auditory characteristics

Spectral band replication (SBR) is an effective tool for low bitrate audio codecs such as HE-AAC and enhanced aacPlus. We propose a new algorithm for the bitrate reduction of SBR, which modifies the envelope data coding of SBR considering the threshold in quiet and the masking threshold. Experimental results show that the proposed algorithm significantly reduces the bitrate compared to the conventional SBR with no degradation in perceived sound quality.

Acoustic Characteristics and Timbre Preferences of Korean Bells

The sounds of the Korean temple bells, that are located in the various places, were recorded and classified into two groups according to the size of bells. The sound preference was investigated with the subjective listening test on the bells of each group. And the acoustic characteristics of the bells such as the frequency, amplitude, beat period, and 20 dB decay rate of the partials was analyzed. The correlation between the acoustic parameters and timbre preference were analyzed and the acoustic characteristics of highly preferred bell sound were presented.

Linear-scale perceptual feature extraction for Speech Bandwidth Extensions

This paper presents a new method to extract linear-scale perceptual feature as a subsitute of MFCCs for highband (3.4kHz∼) in Speech Bandwidth Extensions(BWE). The feature extraction method is based on the mel-scale constrained Nonnegative Matrix Factorization(NMF), which decompose linear-scale log spectrum into a linear combination of mel-scale latent variables. While MFCCs parametrization contains non-invertible procedures, suggested feature is represented in linear-scale and proper to recover the highband time-domain speech. Experiment results report that suggested feature shows better instrumental performance with narrowband MFCCs than real cepstrum without additional computation.

A novel audio stream segmentation method for audio signal discrimination

In this letter, an online segmentation algorithm for audio signal discrimination is presented. By detecting abrupt changes in audio signal features and decimating them followed by strength thresholding, segment boundaries of the audio stream are obtained. The resulting segment boundaries provide efficiency and accuracy for the classification stage of audio signal discrimination system.

Optimizing the sound pressure levels at low frequency limits of electrodynamic loudspeakers

This paper focuses on reproducible low frequency ranges of electrodynamic loudspeaker drivers. The research findings lead to a guide that allows the optimal driver to be selected at the early stages of the design when the consumer electronic products such as home theater systems, digital TVs or PC speakers are being considered. The suggested frequency ranges are based on the acoustics and estimated from statistical data from 1,473 loudspeaker drivers of various sizes. In particular, the lower frequency limit is discussed together with the peak volume displacement which is closely related to the maximum reproducible sound pressure in a loudspeaker.

Psychoacoustic measures of blind audio source separation performance

In this paper, an improved method for evaluating the performance of blind audio source separation (BASS) is discussed. In previous studies, such as described in E. Vincent et al., IEEE Transactions on Speech and Audio Processing, 2006, several computation methods for measuring quality of BASS algorithms e.g., defined by source‐to‐distortion ratio (SDR), source‐to‐interferences ratio (SIR), sources‐to‐noise ratio (SNR) and sources‐to‐artifacts ratio (SAR) are introduced. However, those methods do not take human auditory system into consideration. An improved method is developed by applying preprocessing and using weighted‐inner product in frequency domain instead of simple inner‐product in time domain. The proposed method incorporates well‐known psychoacoustic characteristics e.g., masking effect and equal loudness contours. In comparison with the conventional quality measures, the proposed method shows better correlation with the results of carefully designed listening tests.