Echo suppressor having self-adaptive means

Abstract

Abstract An echo suppressor employed in systems interconnecting a two-wire circuit with a four-wire circuit for substantially eliminating "talker's echo" and for selectively attenuating the calling party's voice signal when the called party at the two-wire line begins speaking. The signal levels of signals on both paths of the four-wire circuit are measured to extract the maximum values of the signals for succeeding time intervals. The extracted value of the outgoing path signal is divided by the extracted value of the incoming path signal to generate an attenuation factor value. This value is multiplied by the value of the voice signal in the incoming path to generate an estimated leakage level signal. When this level is greater than the signal level in the output path (i.e. when there is no outgoing voice signal) the outgoing path is disconnected to prevent a talker's echo signal from being transmitted thereby. When the signal on the outgoing path is greater than the estimated leakage level signal the outgoing path is not disconnected, and an attenuation circuit in the incoming path is activated to attenuate signals in the incoming path. Delay means are provided to prevent the disconnection of the outgoing path at the initial and terminal ends of a voice signal. BACKGROUND OF THE INVENTION Long-distance communication networks are typically comprised of a four-wire circuit wherein the carriers or subcarriers are respectively allotted to outgoing and incoming information signals. A local toll circuit normally connects the four-wire circuit, through terminal equipment, to a calling or called subscriber set which is comprised of the two-wire circuit. A four-wire terminating set (or four-wire to two-wire conversion device) must therefore be provided at the terminal equipment to suitably connect the two-wire and four-wire circuits. The conventional four-wire terminating set is comprised of a hybrid coil and an impedance balancing network for attaining impedance balance between the two and four-wire circuits. However, inasmuch as the impedance of the two-wire circuit varies by an appreciable amount as a function of the distance between the terminal equipment and the local subscriber set at the extremity of the two-wire circuit or according to the performance of telephone exchanges and transmission lines between the terminal equipment and the subscriber set, it can not be expected, as a practical matter, to attain impedance balance for any arbitrary subscriber sets. For the convenience of explanation, it is assumed hereinafter that a calling subscriber at the remote end of the long-distance communication network calls a local subscriber on this side of the network. In the case where the above-mentioned impedance balance is attained, the information signal of the calling (remote) subscriber supplied through the communication network to the four-wire input terminal part of the above-mentioned hybrid coil passes without any ill effect through the hybrid coil and the two-wire circuit to the called (local) subscriber set. In the case where the impedance is not completely balanced, the information signal leaks through the hybrid coil to appear at the output terminal pair of the four-wire circuit thereof, and then is sent back to the calling subscriber, causing the phenomenon commonly referred to as talker's echo. Although the talker's echo produced in a comparatively short distance network does not cause much disturbance in the conversation, the echo produced in a long-distance communication network appreciably affects the quality and performance of the conversation, because the speech of the calling subscriber returns to the person speaking as the talker's echo after a transmission period of the order of several hundred milliseconds. An echo suppressor conventionally used for suppressing or removing the talker's echo has a structure adapted to compare the outgoing signal level at the four-wire output terminal pair of the hybrid coil with a threshold level related to the incoming signal level. so that the outgoing signal circuit can be selectively disconnected to interrupt the outgoing signal level when the result of comparison shows that the outgoing signal level is not higher than the incoming signal level. The threshold level is given a value smaller than the incoming signal level by a prefixed rate. Statistics show that the distribution of echo attenuation factor of hybrids (being the main component for the attenuation in the echo path) throughout the United States has an average of 15 dB with a standard deviation of 3 dB. Therefore, the prefixed rate is selected about 6 dB, thereby to ensure that the threshold level is greater than the leakage (echo) level for most of hybrids (at a 99.7 -percent probability). An example of a conventional echo suppressor of the amplitude comparison type particularly adapted to time-division multiplex PCM information signals is described in detail in the technical report by E. Fariello entitled "A Digital Echo Suppressor for Satellite Circuits" published in the IEEE Transactions on Communications, December 1972. Therefore, further description will not be given here. Generally, there are two mutually exclusive problems that one must deal with in an echo suppressor: reduction of voice clipping of the called subscriber's signal and reduction of echo of calling subscriber's signal. In the case of the conventional echo suppressor described above, the calling subscriber's signal is detected with the threshold level higher than the echo level by 9 dB on the average, so that the reduction of the echo is well attained but the malfunction attributed to the voice clipping appreciably reduces the conversation quality. Especially, the beginning part of a speech having relatively lower level than the remaining part is hardly detected until the level thereof reaches the threshold level set above the outgoing signal level in the state of no called subscriber's signal by 9 dB on the average, resulting in the interruption of the leading part of the speech called "initial clipping". Futhermore, if the attenuation in the echo path is poorer than 6 dB, the threshold level is needed to be adjusted more closely to the incoming signal level itself, bringing about a longer period of initial clipping. The initial clipping is inherent to the conventional echo suppressor of this type that detects the presence of the called subscriber's signal through the level comparison thereof with the threshold level which is not related to the attenuation factor of the echo path associated with the present echo suppressor. BRIEF DESCRIPTION OF THE INVENTION AND OBJECTS It is therefore an object of this invention to provide an echo suppressor which makes it possible to suppress or remove the talker's echo without being affected by the initial clipping. Another object of the instant invention is to provide a novel electronic device for suppressing the talker's echo in a long-distance communication network, coupling a two-wire circuit at the end terminals of a four-wire circuit wherein means are provided for disconnecting the outgoing signal circuit of the four-wire circuit during the presence of the calling subscriber's signal, means for inhibiting the disconnection of the outgoing signal circuit in the presence of the called subscriber's signal detected by the comparison of the outgoing signal level with an estimated leakage (echo) component level. In the present invention, the threshold level for the detection of the called subscriber's outgoing signal is given by the estimated leakage (echo) level generated as follows: A level ratio between the outgoing signal level on the four-wire output terminal pair and incoming signal level on the four-wire input terminal pair is measured on a real time basis, and then multiplied with the incoming signal level to generate the product, i.e., the estimated leakage level. When the called subscriber's signal is not present, the level ratio indicates the real attenuation factor of the leakage (echo) path, and therefore the estimated leakage level is close to the real leakage level. To use the estimated leakage level as the threshold level, a certain delay is given to the attenuation-factor-representing signal, making it possible to avoid the initial clipping. In the presence of the called subscriber's signal, the measured level ratio tends to indicate greater value than the real attenuation factor. The ratio is set to be unity at most. It follows therefore that the sensitivity for the called subscriber's is equal, even in the worst condition, to that of the conventional echo suppressor, where the calling subscriber's signal level itself is used as the threshold level without any attenuation.

Inventors

• 👤 Takashi Araseki
• 👤 Kazuo Ochiai

Bibliographic Details