Earticle

When a person is drunk, lung capacity, lung movement velocity and air pulse change. Drunken people show unique characteristics especially in their speeches with respect to the spectrum area because they tend to speak more forcefully. In other words, the spectrum in the low frequency band area and high frequency band area become relatively flat. This paper suggests that it is possible to develop a sobriety test using speech analysis based on formant energy distribution. The test utilizes the errors in forming the formant envelope between high order linear prediction coding and low order linear prediction coding. About 65% accuracy was obtained in determining whether or not a person is sober by this method.

Unified speech and audio coding (USAC) is one of the latest coding technologies. It is based on a switchable coding structure, and has demonstrated the highest levels of performance for both speech and music contents. In this paper, we propose an extended version of USAC with a single-mode of operation-which does not require a switching system-by extending the linear prediction-coding mode. The main concept of this extension is the adoption of the advantages of frequency-domain coding schemes, such as windowing and transition control. Subjective test results indicate that the proposed scheme covers speech, music, and mixed streams with adequate levels of performance. The obtained quality levels are comparable with those of USAC.

An automatized latency calculation method of evoked otoacoustic emission is proposed. Averaged e-OAE signal is measured from normal adult and it is processed by IBM-AT computer for latency calculation. After partition of stimulated and emitted signal on the time axis by converted linear prediction coding spectrum, latency is calculated by cross correlation method. The output is also compared with those of conventional autocorrelation and pure cross correlation methods. The result show that proposed method has better performance than those of conventional ones.

본 논문은 컴포트 노이즈(comfort noise)를 이용하는 주파수 밴드선택 음성/오디오 코덱에서 컴포트 노이즈 대신 SBR(Spectral Band Replication) 기술을 이용하여 여기 신호를 대체 함으로서 밴드 선택 광대역 음성/오디오 부호화기의 성능 향상을 목표로 한다. 비 전송 밴드에 SBR 기술로 합성된 신호를 삽입하기 위하여 부밴드 별로 전송된 신호를 활용하며, 각각의 부밴드 별로 에너지 가중치를 설정한다. 백색잡음 성분의 컴포트 노이즈 대신 전송신호에 의존하는 신호를 합성 함으로서 보다 높은 음질의 밴드 선택 부호화기를 제안하였다.

This paper is aimed to improve performance of Band-Selection speech/audio Coder reconstucted band spectrum that is not sent by the comfort noise. To improve the performance, we use the Spectral Band Replication(SBR) technique instead of substitution of Comfort noise. To synthesize SBR signal, the SBR algorithm is referenced in selected signals and the spectrum synthesized by SBR is injected to non-selected band. Each sub-band spectrum has been energy-weighted by real audio signal. We propose the enhanced the Band-Selection Coder that utilizes synthesized SBR signal from selected signal instead of comfort noise.