| 摘要 |
For processing a signal applied to an input of audio signal processing devices, for example a speech signal received via a microphone (2) in a speaker-identification or speech recognition system, the signal must have a specific level. For this purpose the signal is applied via an amplifier (4) with variable gain. Before processing begins, the gain is set to an initial value. During processing of the speech signal, it is checked, at regular time intervals, in which range of a plurality of predetermined amplitude ranges the amplified signal is situated, and whether overloading takes place and the frequency which the amplitude of the signal occurs in each of the amplitude ranges is counted. When the signal ends, the optimum gain setting is determined from the histogram of frequencies. In the case of a signal which is specific of an individual speaker and which must be processed in speech recognition or speaker identification systems, this optimum gain setting may be stored in a memory (28) at a location assigned to the speaker, and when the system is subsequently used by the same speaker this value is used as the initial setting of the amplifier gain. In this way the gain is set to an optimum value during the first time that the system is used by a specific speaker and it is corrected continuously when the system is used again by the same speaker. |
