A new technique of pitch extraction has been implemented that makes use of the double spectrum analysis of speech signals. In contrast to previous applications of this method of extracting pitch, the system described here operates in real time, it does not require the use of a computer, important system parameters such as "processing period" can be changed manually or adaptively, and the use of logarithmic compression is optional When logarithmic compression is employed, the pitch extractor represents a real-time implementation of the "cepstrum" technique. The performance characteristics are described and illustrated for practical field conditions, including poor signal-to-noise ratios.