Operation Mode

Figure 4.11 depicts the schematic diagram showing the run-time mode of the implemented tool. Briefly, the original signal is impaired by the encoder and the network. The parameters' values collector gets the values that are fed to the trained NN. The NN evaluates the quality based on the parameters' values. As we see from this Figure, we do not need access to the original signal to evaluate the quality (which is the case for the traditional objective methods depicted by the block diagram in Figure 4.12). This point is very important, as the access to both original and distorted signals constitutes a major disadvantage of the existing objective quality measures which limits their use in real-time applications.

Figure 4.11: The run-time mode of our method.

Figure 4.12: The current existing model of objective methods.

When our method is intended to be used in user's applications to evaluate the quality of the transmitted multimedia in real time, the proposed architecture is as shown in Figure 4.13. At the sending side, the media source (video camera, microphone, or recorded material) generates a continuous media stream. That stream is encoded in real time by the appropriate encoder. In this stage the quality of the signal may be degraded due to the encoder's impairments. The output of the encoder is then packetized and sent to the packet network using a suitable transport protocol (RTP is the most common used one for real-time multimedia transportation in IP networks). Here also, the quality of the signal may be degraded due to network impairments. At the receiving side, the sent stream (stream of packets) is reconstructed by RTP protocol. The encoded signal data is retrieved from the packets (depacketized). Then it is decoded by a suitable decoder. The decoded signal is sent to the appropriate output device (speakers or video card) to be played. The interaction between our tool and the other elements is as follows. The parameters' collector part probes all the working parameters from the encoder, decoder, packet network and the transport protocol. Then the trained NN part evaluates the signal quality as a function of these parameters. In this way, the end-user at the receiver side can see the quality measure instantaneously. While at the sender side, if necessary, the quality can be sent by means of the transport protocol from time to time. (If RTCP protocol is used, it can be sent every 5 sec. However, if other dedicated protocol is used, the quality score can be fed to the sources each time we want.) This means that the frequency update of the parameters and hence the quality evaluation can be done at any time the user wants at the receiver side, while at the sender side the user can have feedback about the quality at least every 5 sec. (again, if RTCP is used).

Figure 4.13: Operation mode for the tool in real-time video system.