Distortion compensation method for speech recognition system - involves speech analysis for data reduction, using division to correct code words

摘要

The method of correcting distortion in speech transmitted from a microphone by lines using the lines for remote control includes carrying out a speech analysis combined with a data reduction of the input signals using the conditions of the analysis for processing. The speech recognition system uses code words at the receiving end which are stored. After forming a detrmined connection between the analysis conditions and the code words, the locally inserted code words are divided, and the results of the division are used for correction, and the corrected analysis parameters are used as a basis for the pattern equalisation. Before or during analysis, a device with logarithmic voltage characteristics such as an A/D converter is used and correction values are obtained by subtracting the analysis parameters of a locally applied code word from those of the speech recognition code word. Correction is made by subtracting the correction values of the analysis parameters along the same path as the speech input path.