| 摘要 |
The subject of the invention is a method of speech compression without quality deterioration, during which the input signal is conducted into a filter bank, where they are separated into partial bands. Later the partial bands are transformed and conducted into a compression unit for comparision with a previously stored signal set. The compressed signal is formed as the result of this transformation and comparison. Its specialty is that after the division of input signals into partial bands, the amplitude of each instantaneous signal is generated, the instantaneous amplitudes are averaged with a low pass filter by partial bands. The segments' averages are logarithmically quantized. The sum of squares of instantaneous amplitudes is filtered with high pass filter, than the period time is defined. Finally the compressed signal is generated by linearly or logarithmically quantizing the period time of segments. |
