5th etsi speech Quality Test Event Anonymous Test Report

Echo Performance - 6 dB ERL

Download 0.63 Mb.

Page	9/13
Date	06.08.2017
Size	0.63 Mb.
	#27239

1 ... 5 6 7 8 9 10 11 12 13

3.7Echo Performance - 6 dB ERL

3.7.1Gateway No.1


Parameter	Reference		Gateway
Echo Loss (G.122)	---		48.3 dB
The echo attenuation is sufficiently high, the measured echo loss is only limited by the idle channel noise level in sending direction. The echo attenuation is further stable vs. time, EC and NLP work reliable for high and low test signal levels (fig. 5.134 and 5.135). Echo components during double talk do not appear if the device is fully adapted before.

Fig. 5.134: Echo att., 5 dB_m0		Fig. 5.135: Echo att., 25 dB_m0			Fig. 5.136: Echo during DT

Fig. 5.137: Simulated DT, enlarged		Fig. 5.138: Echo attenuation			Fig. 5.139: Near end signal, enlarged
The NLP is only inserted under extreme level conditions (fig. 5.138), the near end signal bursts are completely transmitted in fig. 5.139. Clipping does not occur.
					Slight echo components appear in the presence of background noise at the near end. Furthermore NLP inserts some modulation on the transmitted background noise. The Relative Approach analysis detects the echo components (magenta color) and the noise attenuation (dark color).
Fig. 5.140: Noise transmission with far end signal (café)		Fig. 5.141: Δ Relative Approach

3.7.2Gateway No.2


Parameter	Reference		Gateway
Echo Loss (G.122)	---		70.6 dB
The echo loss of 70 dB is very high and indicates a strong influence of NLP. The send signal is completely muted. The echo attenuation is further stable vs. time as shown by the two analyses in figure 5.142 and 5.143. However some echo components can be detected during double talk (fig. 5.144).

Fig. 5.142: Echo att., 5 dB_m0		Fig. 5.143: Echo att., 25 dB_m0			Fig. 5.144: Echo during DT

Fig. 5.145: Simulated DT, enlarged		Fig. 5.146: Echo attenuation			Fig. 5.147: Near end signal, enlarged
The NLP is temporarily inserted under extreme level conditions (fig. 5.146), the near end signal bursts are completely transmitted in fig. 5.147. Clipping does not occur, but the double talk sequence is overlapped by echo components.
					The NLP completely suppresses the near end background noise signal coincident to the application of a far end signal. The background noise features are completely lost. This is indicated by the dark color in the Relative Approach analysis.
Fig. 5.148: Noise transmission with far end signal (café)		Fig. 5.149: Δ Relative Approach

3.7.3Gateway No.3


Parameter	Reference		Gateway
Echo Loss (G.122)	---		48.3 dB
The echo attenuation of 48 dB under this test condition is sufficiently high. The echo is completely suppressed. The echo loss is only limited by the idle channel noise level in sending direction. The echo attenuation is high and constant vs. time (fig. 5.150 and 5.151). Echo components are not detected during the double talk analysis in figure 5.152.

Fig. 5.150: Echo att., 5 dB_m0		Fig. 5.151: Echo att., 25 dB_m0			Fig. 5.152: Echo during DT

Fig. 5.153: Simulated DT, enlarged		Fig. 5.154: Echo attenuation			Fig. 5.155: Near end signal, enlarged
The echo attenuation is high during the double sequence. Echo suppression is inserted under extreme level conditions and further reduces the residual echo after echo cancellation (fig. 5.154). The initial part of the near end signal bursts are clipped, however the duration is too short to cause audible degradation during speech transmission.
					The echo suppression slightly attenuates the transmitted background noise. An adaptive comfort noise masks the attenuation. The Relative Approach analysis detects only a slightly varying signal characteristic in fig. 5.157.
Fig. 5.156: Noise transmission with far end signal (café)		Fig. 5.157: Δ Relative Approach

3.7.4 Gateway No.4


Parameter	Reference		Gateway
Echo Loss (G.122)	---		48.2 dB
The measured echo attenuation is sufficiently high, the result is only limited by the idle noise level. The echo attenuation analysis vs. time fulfills the tolerance for both test signal levels. It is constant and stable vs. time. The echo attenuation is sufficiently also under double talk conditions.

Fig. 5.158: Echo att., 5 dB_m0		Fig. 5.159: Echo att., 25 dB_m0			Fig. 5.160: Echo during DT

Fig. 5.161: Simulated DT, enlarged		Fig. 5.162: Echo attenuation			Fig. 5.163: Near end signal, enlarged
The echo attenuation during double talk is very high, it is again only limited by the idle noise level. The NLP does not insert gaps in the near end signal. All test signal bursts are completely transmitted (fig. 5.161 and 5.163).
					The NLP suppresses the residual echo efficiently. The resulting gaps are masked by an adaptive comfort noise. Although comfort noise can not reproduce the original noise characteristics (café noise) the level is appropriately adjusted and the whole sequence does not attract the attention of the listener.
Fig. 5.164: Noise transmission with far end signal (café)		Fig. 5.165: Δ Relative Approach

3.7.5 Gateway No.5


Parameter	Reference		Gateway
Echo Loss (G.122)	---		48.1 dB
The echo attenuation expressed by the one-dimensional echo loss results is sufficiently high. The echo is completely suppressed in this test applying an average test signal level. Residual echo components occur for high test signal levels (fig. 5.166). Strong echo components can also be detected during double talk in figure 5.168.

Fig. 5.166: Echo att., 5 dB_m0		Fig. 5.167: Echo att., 25 dB_m0			Fig. 5.168: Echo during DT

Fig. 5.169: Simulated DT, enlarged		Fig. 5.170: Echo attenuation			Fig. 5.171: Near end signal, enlarged
Residual echo components appear during the pauses between two near end bursts (fig. 5.169). The echo attenuation during double talk varies due to the insertion of NLP. The initial part of the near signal bursts are clipped during double talk.
					A short term echo component also appears if background noise is present at the near end. Furthermore the NLP attenuates the background noise. The modulation appears as dark colored periods in the Δ Relative Approach analysis.
Fig. 5.172: Noise transmission with far end signal (café)		Fig. 5.173: Δ Relative Approach, Café

Download 0.63 Mb.

Share with your friends:

1 ... 5 6 7 8 9 10 11 12 13

The database is protected by copyright ©ininet.org 2024
send message