Filtering noisy signals


Filtering or suppressing noise in speech transmissions (telephone, mobile communication or other methods of speech transmission)

Communication of people through a speech transmission system is often disturbed by background noise, which complicates evaluation of a surveillance procedure (such as engine noise, music, other sources of disturbance). Audio signals, which are recorded under difficult acoustic conditions, are overlapped by the mentioned disturbance and often faint. The program system tries to localize the disturbing signals and to filter in the frequency domain, to enable a third person to understand the spoken conversation by enhancing the signal to noise ratio. Such calculations are done taking into account the newest psychoacoustic theories. The filtered speech signals are output through a computer in a multi user system or recorded onto a sound storage medium in digital or analog form (see examples). There are situations, where the noise and the speech occupy identical frequency bands and the noise level is much higher than the volume level of the speech. In this case, special filter operations are performed, which enhance understandability a great deal, even in cases which were classified as hopeless. In such cases, filtering can only be done in specific limits though. For situations demanding a live-evaluation, a filter-program has been developed, enabling evaluation to be done in real-time.