Frequency compression makes it possible to transmit intelligible speech in very much less than the usual waveband of 2000?3000 c/s. Two methods are discussed, of which only the first has been experimentally tested. In the first method, frequencies are compressed or expanded by the Doppler effect which arises when a continuously moving record, such as a sound film, is scanned by slits or the like which are themselves running. An experimental device is described which has been constructed for testing the quality of compressed and re-expanded speech. Compression to one-half, followed by re-expansion gives full intelligibility, but some roughness. Speech compressed to one-quarter, and even to one-sixth, remains intelligible, but becomes monotonous, as the inflections of the pitch are not transmitted. The intelligibility is lost only at about eightfold compression. Analysis of the causes of roughness and monotony leads to an improved principle of speech transmission by compression. In the original method only certain ?preferred frequencies? are correctly transmitted, and the reproduction is good only if one of these happens to coincide with the pitch of the speech. In the improved method the preferred frequencies are not fixed but are variable, and adjust themselves automatically to the pitch, which is always correctly transmitted. This also ensures automatically the correct and undistorted transmission of its overtones. Suggestions are made for realizing this principle, which eliminates both roughness and monotony. Frequency-compression systems, however perfect, utilize only one-half of the capacity of a transmission channel, as they do not make use of the information contained in the phase of the spectral components. The concept of ?phase? is analysed, and it is shown that it is possible, at least in principle, to transmit two independent messages through the same channel by a system called ?quadrature transmission? or ?sine-cosine duplex,? but that there are great tec-nhnical difficulties in the way of realizing this system.
展开▼
机译：频率压缩使得可以以比通常的2000？3000 c / s的波段少得多的速度传输可理解的语音。讨论了两种方法，其中只有一种已经过实验测试。在第一种方法中，通过多普勒效应来压缩或扩展频率，该多普勒效应是由连续运行的狭缝等扫描连续的记录（例如声电影）时产生的。描述了一种实验设备，该设备被构造用于测试压缩和重新扩展的语音的质量。压缩到一半，然后再重新膨胀，可以完全理解，但有些粗糙。压缩到四分之一甚至六分之一的语音仍可理解，但由于音调的变化不被传送，因此变得单调。仅在大约八倍压缩时才失去清晰度。分析粗糙和单调的原因导致改进了通过压缩的语音传输原理。在原始方法中，只有某些“首选频率”？正确地传送，并且仅当其中之一恰好与语音的音高一致时，再现才是好。在改进的方法中，优选的频率不是固定的而是可变的，并自动将其自身调整为始终正确传输的音调。这也可以自动确保其泛音的正确且无失真的传输。为实现该原理提出了建议，该原理消除了粗糙度和单调性。频率压缩系统无论多么完美，都仅利用传输信道容量的一半，因为它们没有利用频谱分量相位中包含的信息。 “阶段”的概念进行了分析，结果表明，至少在原则上，可以通过称为“正交传输”的系统通过同一信道传输两个独立的消息。或“正弦余弦双工”？但是在实现该系统的方式上存在很大的技术困难。
展开▼
