Simulation Results

Image transmission over the underwater channel is simulated under different physical conditions to assess the system performance for both unequal and equal error protection with the R-S channel coder. The parameter of wind speed is selected at different values. Each experiment is run 50 times to collect the average. As shown in Fig. 3.9 and Fig. 3.10, where several variations of the standard “Lena” and “dolphin swimming” images are shown [4, 138] following transmission through the proposed underwater communication system. The reconstructed images are obtained at different wind speed levels using the two protection algorithms, unequal and equal error protection.

In the UEP case, the total number of transmitted symbols (R in eq. (3-11)) in the simulation is 67487. The redundancy symbol number for each encoded group is determined by solving the optimization problem defined in eq. (3-11) (the first two packets are significance packets, followed by four sign packets, nine set packets and finally two refinement packets). R-S symbol redundancy is shown in Fig. 3.8 (b) where the significance packets have eight symbol protection level, the sign and set packets have six symbol protection level, and refinement packets are protected by four symbols. In the simulation of the “dolphin swimming” image, the R-S rate distribution changes under various channel wind conditions, as shown in Fig. 3.9 (b). In particular, at 10 Knot wind speed the packets of significance type are protected by eight symbols, six symbols are used, and the refinement packets are completely unprotected. As a result, the total number of transmitted symbols using UEP is 65999, which is 1488 less than that obtained using the traditional EEP technique.

Figure 3.8 shows the average PSNR of the received Lena image transmitted over the UWA channel by using the proposed scheme, MD-Like allocation shame [110], and equal forward error correction. With the R-S coder all coded images are transmitted at certain total bit budget.

Fig. 3.8. The average PSNR of the received Lena image transmitted over the UWA channel for EEP, MD- Allocation [13] and proposed UEP channel coding, at channel Doppler shift = 29.6 Hz.

44

Fig. 3.9. Reconstructed “Lena” image transmitted over the UWA channel with equal and unequal error protection with channel rate distribution of unequal error protection.

Fig. 3.10. Reconstructed “Dolphin Swimming” image transmitted over the UWA channel with equal and unequal error protection with channel rate distribution of unequal error protection.

45

Due to significant UWA channel noise, the curves in Fig. 3.8 are not smooth. The unequal error protection rate allocation encoding method improves the PSNR of the received image. From the simulation results, the proposed UEP method outperforms the conventional EEP system by 3.5 dB and the system of MD-like allocation schemes [110] by 2.2 dB in the underwater acoustic channel with a Doppler shift equal 29.6 Hz at fixed total bit budget. The iterative procedure in the proposed system increases the time needed to encode an image, but results in higher PSNR for reconstructing images in a low SNR channel which makes it more suitable for underwater acoustic channels. Also, from the simulation results, the R-S code is shown to be suitable for an underwater acoustic channel. Please note the equal error protection (EEP) represent the performance when unequal error protection is not used.

In our UEP based on 16-HQAM, the proposed unequal error protection is only for a certain communication system HQAM method, but the rate allocation method can be used with any modulation system. Table 3.4 compares our proposed method and other methods developed elsewhere.

Table 3.4

Comparison of UEP based on Rate Allocation Scheme to other UEP Methods.

Items Proposed scheme MD-Like allocation

scheme [110]. HQAM- approaches

Complexity Complex Complex Simple

CSI (channel state

information) Not required Required Not required

Feedback from receiver to

transmitter Not required Required Not required

Validity for different communication systems

Valid for any communication system

Valid for any communication system

Specific only for HQAM systems

3.6

Conclusion

In this chapter, an efficient unequal error protection technique is proposed for image transmission in the UWA channel. Based on a modified SPIHT coder scheme, a novel image transmission method has been suggested. A SPIHT coder algorithm has been modified based on the order of significance of the coded bits as well as its contribution in the PSNR of the received image. Two UEP methods are proposed, based on 16-HQAM coherent modulation techniques and using rate allocation image transmission. In particular, four different types of bit- streams are generated by a modified SPIHT coder. These four groups are used directly in the rate allocation case, while both groups merge in UEP using HQAM modulation to produce two vectors of bit-streams representing high and low priority bits. In this chapter, 16-HQAM is tested for high-speed image transmission in single and multi- carrier modulation, where 16-HQAM is used as the modulation system and as a mapper of ZP-OFDM. Simulation results show that 16-HQAM is capable to transmitting high-speed images using the proposed underwater communication systems. The 16-HQAM single carrier modulation system is used based on DFE, while the ZP-

46

OFDM uses long guard intervals to avoid the inter-block interference problem. Also, an efficient rate allocation image transmission scheme in underwater acoustic systems is discussed. The R-S channel coder is used together with the output of the rate allocation vectors to generate the output bit-stream for transmission. While such techniques can be considered for generic image transmission issues to achieve reasonable efficiency, an optimization problem is carefully formulated by taking the particular channel feature of underwater acoustic communication channels into account, such as Doppler shift and probability bit error distribution. As a result, noticeable reduction of the overall distortion over an underwater acoustic channel using our proposed technique is observed, together with the improvement in PSNR for the reconstructed image. Simulation experiments are included for illustrative purposes with different image types and different physical parameters such as wind speed.

47

Chapter 4

An Image Transmission Technique Adaptive to Underwater Time-

Dispersive Fading Channels with Progressive Zero Padding OFDM

4.1

Introduction

This chapter proposes a novel image transmission technique to adapt to the particular features of underwater time dispersive fading channels. Such a communication channel is known to have a considerably low data rate and every effort to improve efficiency will contribute to better image transmission. Single-input multi-output orthogonal frequency division multiplexing (SIMO-OFDM) has been applied to achieve low bit error rate (BER) in selective fading channels with significant tap delays [10]. Each transmitted SIMO-OFDM block includes a zero padding (ZP) section of codes working as a guard interval (GI) to avoid inter-block interference (IBI). Effective avoidance of the IBI can be ensured by sufficient GI length, which is usually required to be longer than or equal to the maximum channel tap delay. However, for channels with significant tap delays, such GI may lead to significant bandwidth losses.

It is noted that the data transmission technique over time reversal SIMO-OFDM with an insufficient GI recently published by Liu Zhiqiang and T.C. Yang [10] may potentially be a viable scheme to achieve reasonable bandwidth efficiency. On the other hand, unequal error protection (UEP) techniques for image communication have been shown to be capable of reducing the data rate and achieving high peak signals–to–noise ratio (PSNR) based on the error protection level of the channel coder as reported in [139]. Furthermore, PSNR has been shown to be a useful quality metric for the reconstructed image of lossy compression coders and it is a reasonable approximation to human perception of reconstruction image quality [116]. Combining these techniques, in this chapter UEP is further improved to adapt to the channel IBI level in each transmitted block by a progressive zero padding (PZP) by minimizing IBI under the total GI length constraint.