1. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a1 Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [A first arrival detection method] Date Submitted: [July, 2005] Source: [Yihong Qi, Huan-Bang Li, Shinsuke Hara, Ryuji Kohno, Company: National Institute of Information and Communications Technology ] Contact: Yihong Qi Voice:+81 46 847 5092, E-Mail: yhqi@nict.go.jp] Abstract: [A first arrival detection method is designed to detect weak first arrival signals] Purpose: [To present a first arrival detection method ] Notice:This document has been prepared to assist the IEEE P802.15. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P802.15.

2. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a2 A First-arrival Detection Approach Yihong Qi Huan-Bang Li, Shinsuke Hara and Ryuji Kohno National Institute of Information and Communications Technology (NICT)

3. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a3 Outline A system model Delay estimation for a single-path propagation A first arrival detection method for a multipath environment Simulation results Conclusion

4. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a4 CorrelatorA/D Delay estimation/ First-arrival detection A delay estimate A transmit signal A system model

5. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a5 A Delay Estimation Algorithm

6. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a6 What is the problem? t m+1 tn h(tn) h(tm) h(tm+1) h(tm+Z-1) t m+2 t m+Z Given samples of a correlation function, how to estimate the time instant corresponding to the peak? correlation function

7. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a7 What is information we know? t m+1 tn t m+2 t m+Z correlation autocorrelation correlation = autocorrelation of s(t) +noise The expression is known.Statistics is known. correlation function

8. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a8 How to use the information? Formulate a maximum likelihood estimation. However, it is complicated: One dimension iterative searching Nonlinear autocorrelation function Lots of samples (N) involved

9. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a9 How to simplify? t m+1 tn h(tn) h(tm) h(tm+1) h(tm+Z-1) t m+2 t m+Z Intuition: samples near the peak are more important.  Use less samples Taylor expansion of autocorrelation function around the peak

10. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a10 A simple solution where

11. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a11 A simple solution An algebraic solution, no iterative search Less than 4 samples in general No nonlinear function any more Independent of noise level Optimal in the sense that the estimate is approaching to the theoretical lower limit as over-sampling is sufficiently large.

12. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a12 A First-arrival Detection Method

13. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a13 Why the problem is difficult? The estimation performance would be degraded considerably when the energy of the first arriving component is not dominant among multipath components of a received signal.

14. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a14 Our Approach A similar idea to the solution of near- far problem in multiuser detection. An iterative scheme: In each iteration, the present strongest signal component is estimated, and is removed from sample data to be processed in the next iteration.

15. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a15 h(t) t1t1 t2t2 tntn t h(t 1 ) h(t 2 ) h(t n ) i=1 i=2 i=3 A graphic illustration There are three multipath components, and the second path is strongest.

16. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a16 h(t) t h(t i1+1 ) h(t i1+2 ) A graphic illustration (cont’d) In the first iteration, n=1, we estimate the time delay and amplitude the strongest multipath. Using less samples is better to avoid the interference from other multipath components.

17. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a17 Delay and Amplitude Estimation Delay estimation can use the previous proposed method –Using less samples Amplitude estimation, the ML estimation based on the delay estimate, known autocorrelation function and correlation samples.

18. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a18 h(t) t A graphic illustration (cont’d) The strongest multipath is removed by using the delay estimate, amplitude estimate and the autocorrelation function. Removed

19. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a19 h(t) t n=1 A graphic illustration (cont’d) Since we are only interested on first arrival delay, samples with time instants later than the delay estimated just obtained is removed. Removed

20. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a20 Robust to interference from other strong multipath components Low computational complexity. – let the first arrival component be the M-th strongest component, the complexity is proportional to M. Advantages

21. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a21 Simulation Results

22. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a22 Notation The conventional method: directly select the time instant corresponding to the largest sample. Theoretical limit: assume a continuous correlation function is used. Over-sampling ratio = sampling frequency/effective bandwidth. –E.g., A Gaussian waveform N(0,a^2) has the effective bandwidth 1/2a, a=1ns, effective bandwidth 500MHz.

23. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a23 Example 1: sampling frequency=420M

24. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a24 Example 2: sampling frequency=625M

25. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a25 Example 3: sampling frequency=1.25G

26. July, 2005 Doc: IEEE 15-05-0406-00-004a Qi, Li, Hara, Kohno (NICT) SlideTG4a26 Concluding Remarks A delay estimation method for mitigating the error due to digital sampling A first arrival detection algorithm which is robust to weak first arriving signals