Reception device and signal detection method

Abstract

This reception device comprises: a calculation unit that calculates a likelihood ratio on the basis of one or more reception symbols and a probability density distribution other than a normal distribution; and a threshold determination unit that determines a threshold for detecting a signal on the basis of the calculated likelihood ratio.　

Description

Receiving device and signal detection method

[0001] The present invention relates to a receiving device and a signal detection method.

[0002] In a communication system that transmits information in packets, a receiving device needs to detect the beginning position of a desired signal frame. A commonly used signal detection method is the likelihood ratio test based on the Neyman-Pearson lemma, which can minimize the rate of undetected signals under a certain rate of false positives (see, for example, Non-Patent Document 1). The likelihood ratio test formula shown in Non-Patent Document 1 changes depending on the probability density distribution of noise added to the received signal.

[0003] Conventional terrestrial radio communications using radio waves employ cross-correlation detection, which calculates the cross-correlation between a known signal and a received signal and detects the position where a correlation peak exceeding a preset threshold is confirmed as the desired signal position (see, for example, Non-Patent Document 2). This is equivalent to a likelihood ratio test assuming Gaussian noise distribution as the noise added to the received signal, and is the optimal detection method in Gaussian noise environments such as terrestrial radio. Cross-correlation detection similar to that described above for terrestrial radio has also been commonly used in conventional underwater acoustic communications.

[0004] M. Chiani and MG Martini, “On sequential frame synchronization in AWGN channels”, IEEE Trans. Commun., vol. 54, no. 2, pp. 339-348, Feb. 2006. M. Chitre, J. Potter and S.-H. Ong, “Optimal and Near-Optimal Signal Detection in Snapping Shrimp Dominated Ambient Noise”, IEEE J. Oceanic Eng. vol. 31: 498- 503, Apr. 2006.

[0005] In underwater acoustic communications, impulsive noise that is several tens of dB higher than steady noise is frequently added to received signals. In such communications channels, the probability density distribution of noise deviates significantly from a Gaussian distribution. Therefore, when conventional cross-correlation detection assuming a Gaussian distribution is directly applied to underwater acoustic communications, signal detection performance is significantly degraded compared to when a likelihood ratio test based on the strict probability density distribution of underwater acoustic noise is applied. This performance degradation is particularly noticeable at low SNRs (signal-to-noise ratios), where the effects of impulsive noise become apparent.

[0006] In view of the above circumstances, an object of the present invention is to provide a technology that can improve the performance of signal detection in underwater acoustic communications.

[0007] One aspect of the present invention is a receiving device comprising: a calculation unit that calculates a likelihood ratio based on one or more received symbols and a probability density distribution other than a normal distribution; and a threshold determination unit that performs a threshold determination to detect a signal based on the calculated likelihood ratio.

[0008] One aspect of the present invention is a signal detection method that calculates a likelihood ratio based on one or more received symbols and a probability density distribution other than a normal distribution, and performs a threshold determination for detecting a signal based on the calculated likelihood ratio.

[0009] The present invention makes it possible to improve the performance of signal detection in underwater acoustic communications.

[0010] 1 is a diagram for explaining the characteristics of noise in an underwater acoustic transmission path. FIG. 2 is a diagram showing a Gaussian distribution and a probability density distribution of noise in an underwater acoustic transmission path. FIG. 3 is a diagram showing an example of the configuration of a communication system 100 in a first embodiment. FIG. 4 is a diagram showing an example of the configuration of a signal detection unit 11 in a first embodiment. FIG. 5 is a flowchart showing the flow of a density distribution table creation process performed by a probability distribution estimation unit 13 in a first embodiment. FIG. 6 is a diagram showing an example of the configuration of a density distribution table in a first embodiment. FIG. 7 is a flowchart showing the flow of processing of a receiving device 10 in a first embodiment. FIG. 8 is a diagram showing an example of the configuration of a signal detection unit 11a in a second embodiment. FIG. 9 is a flowchart showing the flow of processing of a receiving device 10a in a second embodiment. FIG. 10 is a diagram showing simulation results in the first and second embodiments.

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0012] (Conventional General Signal Detection Method) First, to facilitate understanding of the embodiments of the present invention, cross-correlation detection, which is a conventional general signal detection method, will be described. Conventional signal detection is performed by statistical hypothesis testing in which the presence or absence of a known signal in a received signal frame is set as a null hypothesis and an alternative hypothesis. Specifically, the null hypothesis H ０ and alternative hypothesis H １ We conduct hypothesis testing based on the above.

[0013]

[0014]

[0015] Here, y in formula (1) and formula (2) ｊ is the received signal vector, and y ｊ = [y ｊ ,y ｊ＋１ ,…,y ｊ＋Ｎ－１ ] Ｔ In equation (2), s is a known signal vector, and s = [s １ ,s ２ ,…,s Ｎ ] Ｔ In formula (1) and formula (2), n ｊis the noise vector, and n ｊ = [n ｊ ,n ｊ＋１ ,…,n ｊ＋Ｎ－１ ] Ｔ That is, the null hypothesis H ０ represents the case where the observed signal does not contain any known signal (only noise), and the alternative hypothesis H １ represents the case where a known signal is included in the observed signal. Based on the Neyman-Pearson lemma, signal detection that minimizes the undetected rate under a certain false positive rate is realized by the likelihood ratio test of the following equation (3).

[0016]

[0017] Here, in equation (3), f ｎ (·) represents the probability density distribution of noise, and λ represents the detection threshold. ｊ When the threshold value λ is exceeded, a signal is detected (decision D １ ), likelihood ratio Λ ｊ is equal to or less than the threshold value λ (determination D ０ ), the received signal vector is advanced by one symbol to the next frame (i.e., j is incremented by 1), and the likelihood ratio calculation continues. In this way, in conventional signal detection, the likelihood ratio is calculated successively while the received signal vector is advanced by one symbol, and the position where the threshold value λ is exceeded is detected as the start position of the desired signal frame.

[0018] As shown in equation (3) above, the likelihood ratio test formula is based on the probability density distribution f ｎ (·), where the probability density distribution of noise f ｎ If a Gaussian distribution is assumed for (·), the likelihood ratio test formula is the cross-correlation detection formula shown in the following formula (4).

[0019]

[0020] Therefore, in a communication channel where noise follows a Gaussian distribution (for example, an AWGN (Additive White Gaussian noise) transmission channel), cross-correlation detection expressed by equation (4) is the optimal signal detection method. For this reason, cross-correlation detection has generally been used as the signal detection method in conventional terrestrial wireless (Gaussian noise environment).

[0021] (Challenges of Signal Detection in Underwater Acoustic Transmission Lines) In underwater acoustic transmission lines, not only stationary noise, such as ocean wave noise and thermal noise from the receiving device, but also impulsive noise, whose amplitude is several tens of dB higher than that of stationary noise, is frequently observed due to the activity of marine organisms such as pistol shrimp, as shown in FIG. 1. The probability density distribution of noise in such an environment, as shown in FIG. 2, has extremely broad tails (e.g., line segment L2 in FIG. 2) compared to a Gaussian distribution (e.g., line segment L1 in FIG. 2), i.e., a distribution that deviates significantly from the Gaussian distribution. Therefore, when conventional cross-correlation detection, which assumes a Gaussian distribution, is directly applied to underwater acoustic transmission lines, there is a problem that detection performance is significantly degraded. Therefore, by performing a likelihood ratio test that takes into account the characteristics of the probability density distribution of underwater acoustic noise, it is possible to improve the performance of signal detection in underwater acoustic transmission lines.

[0022] First Embodiment In the first embodiment, a signal detection method is described in which a likelihood ratio test is performed based on a probability density distribution calculated directly from collected noise data. FIG. 3 is a diagram showing an example configuration of a communication system 100 in the first embodiment. The communication system 100 includes a transmitting device 1 and a receiving device 10. The transmitting device 1 is installed underwater and transmits signals. The transmitting device 1 intermittently transmits signal frames in which a known signal sequence is added to a payload including transmission data. The receiving device 10 intermittently receives the signal frames transmitted from the transmitting device 1.

[0023] The receiving device 10 includes a signal detection unit 11 and a demodulation unit 12. The signal detection unit 11 detects the start position of a signal frame. The demodulation unit 12 performs demodulation processing from the start position of the signal frame detected by the signal detection unit 11, and demodulates the transmitted information.

[0024] 4 is a diagram showing an example of the configuration of the signal detection unit 11 according to the first embodiment. The signal detection unit 11 includes a probability distribution estimation unit 13, an LLR (Log-Likelihood Ratio) calculation unit 14, and a threshold determination unit 15.

[0025] As a pre-processing before starting communication, the probability distribution estimation unit 13 collects noise samples for a predetermined period in the environment where the receiving device 10 is installed, and tabulates the density distribution of the noise to create a density distribution table. Specifically, the probability distribution estimation unit 13 generates a density distribution table by tabulating the density distribution of the noise. １ ,…,N Ｍ ) is input, and a density distribution table is created using the input noise sequence.

[0026] The LLR calculation unit 14 calculates an LLR value for each received symbol included in the received symbol sequence based on the received symbol sequence and the density distribution table created by the probability distribution estimation unit 13. The LLR calculation unit 14 calculates an LLR value for each received symbol included in the received symbol sequence using the density distribution table created by the probability distribution estimation unit 13. That is, the LLR calculation unit 14 in the first embodiment calculates the LLR value based on a probability density distribution other than a normal distribution. The normal distribution here means the probability density distribution of white Gaussian noise. The LLR calculation unit 14 is one aspect of a calculation unit.

[0027] The threshold decision unit 15 detects the signal by making a threshold decision based on the LLR value for each received symbol calculated by the LLR calculation unit 14. The method of signal detection performed by the threshold decision unit 15 is the same as in the conventional method.

[0028] 5 is a flowchart showing the flow of the density distribution table creation process performed by the probability distribution estimation unit 13 in the first embodiment. The process in FIG. 5 is executed before the receiving device 10 starts communication. After the receiving device 10 has executed the density distribution table creation process once, it is not necessary to execute it every time before starting communication, and it may execute it at a predetermined timing, such as when an instruction is given by the user.

[0029] The probability distribution estimation unit 13 collects a predetermined number of noise samples (step S101). At this time, the receiving device 10 does not transmit a signal. The probability distribution estimation unit 13 calculates the value of the collected noise samples as P １ ,…,P ＬL is an integer equal to or greater than 2. As shown in FIG. 2, the noise distribution in the ocean has an extremely wide tail, so many noise samples are required. For example, the probability distribution estimation unit 13 estimates 10 ６ This corresponds to several minutes (approximately 10 minutes at most) of noise observation in underwater acoustic communications.

[0030] The probability distribution estimation unit 13 estimates the noise sample values ​​P １ ,…,P Ｌ Using the cumulative distribution f ｐ Specifically, the probability distribution estimation unit 13 first calculates (x) from a predetermined number of collected noise sample values ​​P １ ,…,P Ｌ Using the above, the average μ ｐ Furthermore, the probability distribution estimation unit 13 calculates the predetermined number of collected noise sample values ​​P １ ,…,P Ｌ Using the above, the dispersion parameter σ is calculated based on the following equation (6): ｐ In equations (5) and (6), L represents the total number of noise samples, and P ｌ represents the value of the l-th noise sample (l is an integer equal to or greater than 1).

[0031]

[0032]

[0033] Thereafter, the probability distribution estimation unit 13 calculates the calculated mean μ ｐ and the dispersion parameter σ ｐ and the noise sample value P １ ,…,P Ｌ , the noise sample value P is calculated based on the following equation (7): １ ,…,P Ｌ The average μ ｐ 0, and the dispersion parameter σ ｐ is normalized to 1.

[0034]

[0035] P in formula (7) ｊ ´represents a normalized value of the l-th noise sample value. Then, the probability distribution estimation unit 13 creates a table of cumulative distribution (empirical cumulative distribution of noise) based on the following equation (8).

[0036]

[0037] In equation (8), #{P ｌ ´ <x|l=1,...,L} represents the number of noise samples smaller than x. Then, the probability distribution estimation unit 13 calculates the noise density distribution f ｐ (x ｎ ) is calculated (step S103).

[0038]

[0039] The probability distribution estimation unit 13 calculates the density distribution f of the noise. ｐ (x ｎ ) to change the value of n, the noise density distribution f ｐ (x ｎ ) in a table as shown in FIG. 6, thereby tabulating the noise density distribution (step S104). FIG. 6 is a diagram showing an example of the configuration of a density distribution table in the first embodiment. The noise density distribution stored in the density distribution table is created based on noise samples measured in advance in the environment in which the receiving device 10 is installed, as described above. Therefore, the noise density distribution is a probability density distribution that is closer to the density distribution of noise occurring in the environment in which the receiving device 10 is installed, compared to the noise density distribution obtained assuming an AWGN transmission path.

[0040] 7 is a flowchart showing the flow of processing by the receiving device 10 in the first embodiment. Note that the processing in FIG. 7 is processing during actual communication between the transmitting device 1 and the receiving device 10, and it is assumed that a density distribution table has already been created.

[0041] The LLR calculation unit 14 calculates the received symbol sequence (y ｊ ,y ｊ＋１, …,y ｊ＋Ｎ－１ ) (step S201). The LLR calculation unit 14 calculates the collected received symbol sequence (y ｊ ,yｊ＋１, …,y ｊ＋Ｎ－１ ) and the density distribution table created by the probability distribution estimation unit 13, and calculates an LLR value for each received symbol included in the received symbol sequence based on the following equation (10) (step S202). Specifically, the LLR calculation unit 14 calculates an LLR value for each received symbol included in the received symbol sequence based on the numerator of equation (10) by adding a desired known signal s ｎ In other words, if we assume that the noise component x ｎ y ｊ＋ｎ－１ -σ ｓ s ｎ 8, and in the denominator of equation (10), the desired known signal s is added to the received signal. ｎ is not included, that is, the noise component x ｎ y ｊ＋ｎ－１ The density distribution table shown in FIG. 6 is referenced.

[0042]

[0043] Note that the function f in equation (10) ｐ When calculating (x), if x is not included in the density distribution table of FIG. 6, the column corresponding to the nearest neighbor of x is searched for in the density distribution table of FIG. 6, and the value of the column corresponding to the nearest point is calculated as f ｐ Alternatively, the density distribution table in FIG. 6 may be interpolated to obtain the f corresponding to x. ｐ (x) may be estimated.

[0044] The LLR calculation unit 14 calculates the LLR value Λ from the calculated LLR value for each received symbol based on equation (10). ｊ The threshold decision unit 15 calculates the LLR value Λ output from the LLR calculation unit 14 and outputs it to the threshold decision unit 15 (step S203). ｊ The threshold decision unit 15 receives the input LLR value Λ ｊ (Step S204). Specifically, the threshold decision unit 15 performs threshold decision based on the LLR value Λ ｊ The threshold decision unit 15 compares the LLR value Λ with a preset threshold value to make a threshold decision. ｊ It is determined whether or not the value exceeds the threshold value (step S205).

[0045] The threshold determination unit 15 determines the LLR value Λ ｊ exceeds the threshold (step S205-YES), the threshold decision unit 15 detects the position of j as the head position of the received signal. ｊ ,y ｊ＋１, …,y ｊ＋Ｋ－１ ) to the demodulator 12. The demodulator 12 receives the received symbol sequence (y ｊ ,y ｊ＋１, …,y ｊ＋Ｋ―１ ) is demodulated (step S206).

[0046] On the other hand, the threshold determination unit 15 determines the LLR value Λ ｊ If it is determined that j does not exceed the threshold (step S205-NO), the threshold determination unit 15 advances the received signal vector by one symbol to the next frame (step S207). That is, the threshold determination unit 15 advances the received signal vector by one symbol to the next frame by incrementing the value of j by 1. Then, the process returns to step S201 and signal detection continues.

[0047] The receiving device 10 configured as described above can improve the performance of signal detection in underwater acoustic communications. Specifically, the receiving device 10 calculates an LLR value Λ based on a received symbol sequence including one or more received symbols and a probability density distribution other than a normal distribution. ｊ an LLR calculation unit 14 that calculates the calculated LLR value Λ ｊ and a threshold determination unit 15 that performs threshold determination for signal detection based on the probability density distribution other than the normal distribution. Here, the receiving device 10 calculates the LLR value for each received symbol using a density distribution table created based on noise samples measured in advance in the environment where the receiving device 10 is installed, as a probability density distribution other than the normal distribution. In this way, the receiving device 10 calculates the LLR value for each received symbol using a probability density distribution that is close to the density distribution of noise occurring in the environment where the receiving device 10 is installed. Therefore, the estimation accuracy of the LLR value for each received symbol can be improved. As a result, it is possible to improve the performance of signal detection in underwater acoustic communication.

[0048] The receiving device 10 creates a density distribution table using the noise distribution obtained by noise measurement before actual communication, and calculates the LLR value for each received symbol using the created density distribution table. In this way, the receiving device 10 collects noise generated in the environment in which the receiving device 10 is installed and creates a density distribution table in advance before communication begins. This makes it possible to calculate the LLR value for each received symbol using a density distribution table that stores a probability density distribution that is close to the density distribution of noise generated in the environment in which the receiving device 10 is installed. Therefore, the estimation accuracy of the LLR value for each received symbol can be improved. As a result, it is possible to improve the performance of signal detection in underwater acoustic communication.

[0049] (First Modification of the First Embodiment) The probability density distribution of noise may be generated by other methods than the method shown in Fig. 5. Furthermore, if the probability density distribution of noise is known, it does not need to be estimated.

[0050] (Second Modification of the First Embodiment) In the above-described embodiment, the receiving device 10 creates the density distribution table, but the receiving device 10 may acquire the density distribution table from an external device. In such a configuration, the LLR calculation unit 14 of the receiving device 10 acquires the density distribution table created in advance in an external device and calculates the LLR value for each received symbol.

[0051] Second Embodiment In the first embodiment, in order to accurately directly calculate the probability density distribution of noise, it was necessary to measure noise in advance for a long period of time in the environment where the receiving device is installed before starting communication. In the second embodiment, a configuration will be described that can approximate the noise density distribution by measuring noise in advance for a short period of time. Specifically, in the second embodiment, the density distribution of underwater noise with a wide tail is approximated by a t-distribution, and then the signal power is estimated and a likelihood ratio test is performed.

[0052] 8 is a diagram showing an example of the configuration of a signal detector 11a in the second embodiment. The signal detector 11a includes an LLR calculator 14a, a threshold determiner 15, and a parameter estimator 16. The signal detector 11a differs in configuration from the signal detector 11 in that it does not include the probability distribution estimator 13, that it includes an LLR calculator 14a instead of the LLR calculator 14, and that it newly includes a parameter estimator 16. The following description will focus on the differences from the signal detector 11.

[0053] The parameter estimation unit 16 estimates a noise sequence (N １ ,…,N Ｍ ) is input, and the dispersion parameter σ is estimated based on the input noise sequence according to the following equation (11) or (12). The parameter estimation unit 16 also estimates a parameter v representing the width of the tail of the distribution based on the estimated dispersion parameter σ and equation (13).

[0054]

[0055]

[0056]

[0057] f in equation (13) ｔ (N ｊ ;0,ν ｉ , σ) is the density function of the t-distribution, and is expressed as in the following equation (14): Γ in equation (14) represents the gamma function.

[0058]

[0059] The LLR calculation unit 14a calculates the received signal power σ from the received symbol sequence based on the least squares criterion. ｓ Furthermore, the LLR calculation unit 14a estimates the t-distribution parameters (σ, ν) and the estimated received signal power σ ｓ The LLR calculation unit 14a calculates an LLR value for each received symbol included in the received symbol sequence based on the t-distribution parameters (σ, ν) estimated by the parameter estimation unit 16. That is, the LLR calculation unit 14a in the second embodiment calculates the LLR value based on a probability density distribution other than a normal distribution.

[0060] 9 is a flowchart showing the flow of processing performed by the receiving device 10a in the second embodiment. The parameter estimation unit 16 estimates a noise sequence (N １ ,…,N Ｍ ) (step S301). The parameter estimation unit 16 calculates the noise sequence (N １ ,…,N Ｍ ) to estimate t-distribution parameters (σ,ν) based on the above equations (11) or (12) and (13) (step S302). Then, the parameter estimation unit 16 approximates the noise density distribution with a t-distribution using the estimated t-distribution parameters (σ,ν) (step S303).

[0061] The parameter estimation unit 16 outputs the variance parameter σ and the estimated tail width ν of the density distribution as t-distribution parameters (σ, ν) to the LLR calculation unit 14a. The LLR calculation unit 14a receives the t-distribution parameters (σ, ν) output from the parameter estimation unit 16. Furthermore, the LLR calculation unit 14a calculates the t-distribution parameters (σ, ν) of the received symbol sequence (y ｊ ,y ｊ＋１, …,y ｊ＋Ｎ－１ The LLR calculation unit 14a calculates the LLR based on the input t-distribution parameters (σ, ν) and the collected received symbol sequence (y ｊ ,y ｊ＋１, …,y ｊ＋Ｎ―１ ) and the signal power σ is calculated based on the least squares criterion as shown in the following equation (15). s The maximum likelihood estimate of is calculated (step S304).

[0062]

[0063] Here, |y in equation (15) ｊ -σ s s│ ２ Since is a downward convex function, the signal power σ s The maximum likelihood estimate of is given by the following equation (16):

[0064]

[0065] The LLR calculation unit 14a calculates the parameters (σ, ν, σ) obtained in steps S302 and S304. ｓ ) to calculate the LLR value Λ based on the following equation (17): ｊ is calculated (step S305).

[0066]

[0067] The LLR calculation unit 14a calculates the LLR value Λ ｊ The threshold decision unit 15 outputs the LLR value Λ output from the LLR calculation unit 14a to the threshold decision unit 15 (step S306). ｊ The threshold decision unit 15 receives the input LLR value Λ ｊ (Step S307). Specifically, the threshold decision unit 15 performs threshold decision based on the LLR value Λ ｊ The threshold decision unit 15 compares the LLR value Λ with a preset threshold value to make a threshold decision. ｊ It is determined whether or not the value exceeds the threshold value (step S308).

[0068] The threshold determination unit 15 determines the LLR value Λ ｊ exceeds the threshold (step S308-YES), the threshold decision unit 15 detects the position of j as the start position of the received signal. ｊ ,y ｊ＋１, …,y ｊ＋Ｋ－１ ) to the demodulator 12. The demodulator 12 receives the received symbol sequence (y ｊ ,y ｊ＋１, …,y ｊ＋Ｋ―１ ) is demodulated (step S309).

[0069] On the other hand, the threshold determination unit 15 determines the LLR value Λ ｊ If it is determined that j does not exceed the threshold (step S2307—NO), the threshold determination unit 15 advances the received signal vector by one symbol to the next frame (step S310). That is, the threshold determination unit 15 advances the received signal vector by one symbol to the next frame by incrementing the value of j by 1. Then, the process returns to step S201 and signal detection continues.

[0070] According to the receiving device 10 of the second embodiment configured as above, it is possible to obtain the same effects as those of the first embodiment.

[0071] The receiving device 10 in the second embodiment approximates the noise density distribution using a t-distribution. This approximation using a t-distribution can be achieved by short-term noise measurement, and the receiving device 10a does not need to measure noise for a long period of time before starting communication, as in the first embodiment. Therefore, communication can be started immediately without prior long-term noise measurement.

[0072] (Effectiveness evaluation of the proposed signal detection method through computer simulation) We evaluated the effectiveness of the proposed optimal signal detection method, which directly calculates the noise probability density distribution from noise samples, and the proposed signal detection method, which approximates the noise's t-distribution and estimates the signal power, through numerical simulation. The comparison was performed with conventional cross-correlation detection. The parameters of the t-distribution in the proposed method were estimated from 2,000 samples of noise data.

[0073] The simulation conditions were as follows: Noise model: Generated from the probability density distribution of noise measured in actual sea areas (the L2 distribution in Figure 2) Known signal sequence: M sequence (127 symbols) No-signal section: 1000 symbols before and after the known signal section Threshold λ: Optimal value determined from the ROC (Receiver Operating Characteristic) curve Signal detection techniques: The technique proposed in the first embodiment (optimal signal detection in which the noise probability density distribution is calculated directly from the noise sample and the signal power is known), the technique proposed in the second embodiment (t-distribution approximation of the noise probability density distribution + signal power estimation), and the conventional technique (cross-correlation detection).

[0074] The results of a simulation performed under the above simulation conditions are shown in Figure 10. Figure 10 shows the simulation results for a shallow water area (corresponding to the distribution of L2 in Figure 2) where the noise probability density distribution deviates significantly from a Gaussian distribution. In Figure 10, the horizontal axis represents the SNR, and the vertical axis represents the false detection rate. Furthermore, in Figure 10, line segment L5 represents the simulation results obtained using the conventional technology (cross-correlation detection), line segment L6 represents the simulation results obtained using optimal signal detection, which is the technology proposed in the first embodiment, and line segment L7 represents the simulation results obtained using the technology proposed in the second embodiment.

[0075] As shown in Figure 10, we confirmed that the proposed technique can achieve performance close to optimal signal detection under the practical condition of observing a small number of noise samples. We confirmed that it can achieve a performance improvement of approximately 5.5 dB compared to conventional cross-correlation detection.

[0076] (Modification of the Second Embodiment) The t-distribution parameters (σ,ν) in the above formulas (11), (12), and (13) may be fixed values. In this configuration, the receiving device 10 does not need to include the parameter estimation unit 16.

[0077] Some or all of the functional units of the receiving device 10 in the first and second embodiments are realized as software by a processor such as a CPU (Central Processing Unit) executing a program stored in a storage device having a non-volatile storage medium (non-transitory storage medium) and a storage unit. The program may be recorded on a computer-readable non-transitory storage medium. Examples of computer-readable non-transitory storage media include portable media such as flexible disks, magneto-optical disks, ROMs (Read Only Memory), and CD-ROMs (Compact Disc Read Only Memory), and storage devices such as hard disks built into computer systems.

[0078] Some or all of the functional units of the receiving device 10 in the first and second embodiments may be realized using hardware including an electronic circuit (electronic circuit or circuitry) using, for example, an LSI (Large Scale Integrated circuit), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array).

[0079] Although an embodiment of the present invention has been described above in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs within the scope of the gist of the present invention.

[0080] The present invention is applicable to underwater acoustic communications.

[0081] REFERENCE SIGNS LIST 1... transmitting device, 10, 10a... receiving device, 11... signal detection unit, 12... demodulation unit, 13... probability distribution estimation unit, 14, 14a... LLR calculation unit, 15... threshold determination unit, 16... parameter estimation unit

Claims

1. A receiving device comprising: a calculation unit that calculates a likelihood ratio based on one or more received symbols and a probability density distribution other than a normal distribution; and a threshold determination unit that performs threshold determination to detect a signal based on the calculated likelihood ratio.

2. The receiving device according to claim 1, wherein the calculation unit calculates the likelihood ratio based on a probability density distribution of noise obtained by noise measurement.

3. The receiving device according to claim 2, wherein the noise probability density distribution is generated based on noise samples collected in an environment in which the receiving device is installed before communication starts.

4. The receiving device according to claim 2 or 3, further comprising a probability distribution estimating unit that estimates a probability density distribution of the noise based on one or more noise sequences corresponding to the one or more received symbols.

5. The receiving device according to claim 1, wherein the calculation unit calculates the likelihood ratio by approximating a probability density distribution other than the normal distribution based on a t-distribution and estimating a signal power.

6. The receiving device according to claim 5, wherein the calculation unit estimates the signal power based on a least squares criterion.

7. The receiving device according to claim 5 or 6, further comprising a parameter estimation unit that estimates a t-distribution parameter based on one or more noises corresponding to the one or more received symbols, wherein the calculation unit uses the one or more received symbols and the t-distribution parameter to approximate a probability density distribution other than the normal distribution based on a t-distribution, and estimates the signal power based on a least squares criterion.

8. A signal detection method comprising: calculating a likelihood ratio based on one or more received symbols and a probability density distribution other than a normal distribution; and performing a threshold decision for detecting a signal based on the calculated likelihood ratio.

Images