=========================================================================== VLCS 2017 Review #4A --------------------------------------------------------------------------- #4 SeedLight: hardening LED-to-Camera communication with RLC --------------------------------------------------------------------------- Overall merit: C. Weak paper, though I will not fight strongly against it Novelty: 3. New contribution Writing: 2. Needs improvement Reviewer expertise: 3. Knowledgeable ===== Paper summary ===== This paper proposes to use pseudo random linear coding to achieve efficient redundancy for low-end LED-to-Camera communication. For the proposed coding scheme, the code overhead and complexity are reduced in the design. A testbed is built to evaluate the proposed scheme and several key coding parameters are exploited. Overall, the proposed scheme can improve the system throughput/Goodput. However, the key part (Section 3) is not presented clearly and not easy to follow. The whole paper still needs a lot of improvements: some details are missing; results are poorly explained, etc. ===== Comments for author ===== Strength: Apply random linear coding to LED-to-Camera link is worth exploiting. The proposed coding scheme is implemented and experimentally evaluated. The system Goodput can be improved. Weakness: Section 3.2: The proposed pseudo-random generator scheme is not presented clearly in Section 3.2. Since the mapping between the ss and the NN coefficients c_1,…,c_N is the key part, more details should be provided instead of just one sentence. If this mapping is totally borrowed from [16], then the novelty of the proposed scheme is very limited. Section 4.4: the explanations of the results are very casual. “At this distance, the ambient light starts interfering with the LED signal.” Did you measure the ambient light during the experiments? There is no information about the ambient light. “at 5cm, the Goodput with SeedLight is 2.5kbps…. 70%”: From Fig. 3, the Goodput IS 2.25kbps (NOT 2.5kbps) at 5cm, therefore, the gain is much less than 70%. “a gain of 100% is noticed at distance of 30cm”: it is very clear from the figure that gain is less than 100%. The authors should give accurate numbers, not a “casual” number. Other comments: Eq.(1): the operator TIMES should be placed in the middle, not a dot. Similarly in page 5, left column, bottom. The quality of the figures should be improved. Not readable when printing the paper out. Page 5, left column, “(3) requires the largest…”: use “M3” or “Method3” instead “(3)”. Page 6, left column, “SeedLight hence reduces the energy consumed”. No results to support this claim. Although the retransmission is reduced, you need overhead on the random linear coding. Page 3, right column, “up to to 30 frames” -> “up to 30 frames” Page 3, right column, “following Fig.1.” -> “following the structure shown in…” =========================================================================== VLCS 2017 Review #4B --------------------------------------------------------------------------- #4 SeedLight: hardening LED-to-Camera communication with RLC --------------------------------------------------------------------------- Overall merit: B. OK paper, but I will not champion it Novelty: 2. Incremental improvement Writing: 3. Adequate Reviewer expertise: 2. Some familiarity ===== Paper summary ===== The paper presents a coding scheme that attempts to deal with inherent packet losses of LED-Camera communication and aims to enhance line-of-sight LED-to-Camera communication goodput. The scheme has been implemented and a preliminary evaluation using the camera of an Android phone is included. ===== Comments for author ===== Please state the version of Android you use, as well as if you made modifications (or what modules you modified). The evaluation requires more work be convincing. You should use at least 2 different Android phones; you should discuss any timing dependences and how you deal with them. =========================================================================== VLCS 2017 Review #4C --------------------------------------------------------------------------- Paper #4 SeedLight: hardening LED-to-Camera communication with RLC --------------------------------------------------------------------------- Overall merit: C. Weak paper, though I will not fight strongly against it Novelty: 2. Incremental improvement Writing: 3. Adequate Reviewer expertise: 2. Some familiarity ===== Paper summary ===== Many of the previously proposed light-to-camera communication systems use coding for robust transmission. This paper proposes a scheme that avoids transmitting the coding coefficients, in order to reduce the overhead and improves the goodput. To do so, it utilizes a pseudo random sequence and only transmits the seed. The coefficients are selected from the sequence and thus the receiver can reconstruct them from the received seed. Evaluation results show that in the best case (where the distance is less than 5 cm) goodput can be improved from 1.5 Kbps to 2.5 Kbps. ===== Comments for author ===== The main motivation of the work seems to be lowering the transmission overhead (to improve the throughput) and to reduce the required computation (to use it on systems with cheap MCUs). But for me, these are not very strong motivations, considering the applications. The difference between 1.5 Kbps and 2.5 Kbps at close distance does not seem to make a big difference (what application works at 2.5 Kbps but not at 1.5 Kbps?), and the difference is even smaller at longer distance (>20 cm, see figure 3). While I believe what is proposed in the paper can achieve the stated objective (and evaluation results provide the proofs), it seems to me it is more important to address the issues in th following, so that it can support meaningful applications. (1) Currently the system only support very short distance (up to 15-20cm) - this limits the applications that can be supported. Currently it looks like a NFC system. (2) Also - it seems the scheme is tailored toward image sensors with the same rolling shutter sampling rate. (Same as the original work by Danakis et al.) Using a different smartphone then the system would likely fail. From the paper, it seems that transmission of seed is not protected by additional coding and would have the same probability of being lost in IFG compared to other transmitted bytes. Would it make sense to employ additional coding to improve robustness? If seed is lost then all subsequent transmitted blocks cannot be decoded, yes? Minor points: the figures in the paper are way too small to read. The labels can hardly be seen. =========================================================================== VLCS 2017 Review #4D --------------------------------------------------------------------------- #4 SeedLight: hardening LED-to-Camera communication with RLC --------------------------------------------------------------------------- Overall merit: B. OK paper, but I will not champion it Novelty: 2. Incremental improvement Writing: 3. Adequate Reviewer expertise: 3. Knowledgeable ===== Paper summary ===== The paper presents a new coding scheme for LED-to-camera communication to improve system goodput. It targets low-power LEDs and low-cost MCUs, and thus focuses on making the design lightweight. ===== Comments for author ===== The work advances a prior work [5] and improves the throughput from 1.5 kbps to 2.5 kbps. It is a solid work. The contribution seems quite incremental though. Given the rich literature, it'd be nice to compare SeedLight to other LED-camera communication schemes, in addition to [5]. As shown in Fig. 2(a), the error probability is generally at least 30% - 60%, which is very high for data communication. It's unclear what applications can be supported by such performance. Fig. 3(a): why does 6-bit seed lead to goodput lower than that of 8 bit? A color LED is used as the transmitter, would LED color affect the performance at all? Figures in the paper are generally hard to read. Font is too small.