Abstract:
Provided is a method for generating metadata for symbolic music and a music restoring apparatus employing the same, and more particularly, a method for generating metadata for ornaments existing in the Korean traditional music and an apparatus for restoring multimedia including the same. A method for generating metadata for music, which is suggested in the present invention, includes the steps of: a) generating an extensible Markup Language (XML) notation for ornaments; b) generating a complete form of metadata including the XML notation; and c) generating metadata for entire music by using the complete form of metadata and the XML notation.
Abstract:
PURPOSE: A method for recovering packet loss by using additional voice data is provided to use voice packet data as additional information for an FEC(Forward Error Correction), and to use additional voice packet data for an error correction in case a packet loss occurs, then to use an error hiding method in case the packet loss consecutively occurs, thereby improving deteriorated sound quality caused by the packet loss. CONSTITUTION: A voice signal is inputted in 1 frame unit(S61). A system decides whether voice signals of all frames are inputted(S62). If not, the system starts coding(S63). The coding process is divided into a low bit rate coding(S64) and a high bit rate coding(S66). The low bit rate coding is performed at 6.6kbit/s bit rate, and a voice signal coded at low bit rate is delayed for 1 frame(S65). A voice signal coded at high bit rate and the voice signal coded at the low bit rate are synthesized into one packet(S67). If the voice signals of all the frames are inputted, the system transmits the packets generated before this time(S68).
Abstract:
PURPOSE: A voice activity detection apparatus and a method thereof in a mobile communication system are provided to distinguish a voice section from a background noise section efficiently while taking over the telephone existing the background noise by using peak measurement(PM) of a residual signal of an LPC(Linear Prediction Coefficient). CONSTITUTION: An LPC analyzer(101) receives and analyzes an input voice signal applied, and then extracts an LPC. An LPC mixer(102) reproduces the voice signal from the LPC. An adder(103) adds up the input voice signal and deducts the signal received from the LPC mixer(102) to produce an LPC residual signal. A PM calculating device(104) performs a PM by using the LPC residual signal. A first flag determining device(105) determines a flag by using the value of PM received from the PM calculating device(104). An analysis and flag determining unit analyzes frame energy and characteristic and distribution of frequency and then determines a flag. A VAD(Voice Activity Detection) device(112) determines a VAD by using a flag received from the analysis and flag determining unit.
Abstract:
본 발명은 하모닉 모델을 사용하는 유/무성음 혼합신호의 하모닉 노이즈 음성 부호화기 및 부호화 방법에 관한 것으로서, 입력되는 LPC 잔여신호를 캡스트럼을 이용하여 무성음 성분인 노이즈를 분리한 후 LPC 분석법으로 스펙트럴을 예측하여 상기 노이즈를 부호화 하는 노이즈-스펙트럴 추정 수단을 포함하는 것을 특징으로 하며, 유/무성음 혼합 신호를 기존의 하모닉 모델에 캡스트럼-LPC 분석법을 통해 예측된 노이즈 스펙트럴 모델을 사용하여 효과적으로 노이즈 분석을 하여 부호화 함으로써, 보다 개선된 음질을 구현할 수 있는 것을 특징으로 한다.
Abstract:
본 명세서는 전력 스펙트럼 특성을 이용하여 스펙트럼을 감지하는 방법 및 장치에 관한 것이다. 이러한 본 명세서는 신호를 수신하는 단계, 상기 수신된 신호를 복수의 데이터 세트로 분할하는 단계, 상기 분할된 각각의 데이터 세트에 대해 피리오도그램을 계산하는 단계, 상기 피리오도그램에 대한 전력 스펙트럼을 계산하는 단계 및 상기 전력 스펙트럼의 최대값과 최소값의 비를 미리 설정된 임계값과 비교하여 우선 사용자가 해당 채널을 사용하는지 여부를 판단하는 단계를 포함하는 스펙트럼 감지 방법을 제공한다.