论文部分内容阅读
最近:美国MITRE公司试制出一种运送连结波锋和波谷之矢量的16kb/s话音编码设备。它传送话音波形锋和谷的时间位置及幅度。此方式称Critical Point Coding,已经用MAP300阵列处理器作了实时处理实验,再生音质良好。编码规程如下:把输入话音带宽限制在3600Hz8kHz取样,12位A/D变换。然后用TDHS(Time DomainHarmonic Scaling)在时域内压缩信号的频带(压缩比为2)。再调查所得波形的斜率,求得极性有变化的点(峰和谷)之位置。还要求得连结一个点与下一个点的矢量(相连续的峰和谷的时间差及幅度差)。时间差用3比特量化,幅度差用8比特量化。用536比特构成一帧。一帧由同步信号、基音信息和44个矢量组成。解码时,利用逆处理得到的峰和谷按3次多项式内插,获得锋和谷之间的数据。
Recently: MITER Corporation in the United States prototyped a 16kb / s voice encoding device that carries a vector of wave front and trough. It transmits the location and amplitude of the voice waveform front and valley. This approach called Critical Point Coding, MAP300 array processors have been used for real-time processing experiments, good reproduction quality. Coding procedures are as follows: the input voice bandwidth is limited to 3600Hz8kHz sampling, 12-bit A / D conversion. The band of the signal is then compressed in the time domain using TDHS (Time Domain Harmonic Scaling) (compression ratio of 2). The slope of the resulting waveform is then investigated to find the location of the point (peak and valley) where the polarity changes. It is also required to concatenate the vector of one point and the next (the time difference and amplitude difference between successive peaks and valleys). The time difference is quantized with 3 bits, and the amplitude difference is quantized with 8 bits. Make a frame with 536 bits. A frame by the synchronization signal, pitch information and 44 vectors. Decoding, the use of inverse-processed peaks and valleys by 3 polynomial interpolation to obtain the data between the front and the valley.