Optical signal processing device

Abstract

There is provided an optical signal processing device capable of RC in a complex space using optical intensity and phase information. An optical modulator controlled by an electric signal processing circuit modulates laser light, which is emitted from a laser light source, at a modulation period either or both of the intensity and phase values of the optical electric field. On the other hand, an input signal is also modulated by the optical modulator at a modulation period in the time domain so as to be an input signal. The converted input signal passes through an optical transmission path and enters an optical circulation circuit via an optical coupler. Part of the circulating light is branched into two by an optical coupler, and the branched light is converted into a complex intermediate signal at a coherent optical receiver. This complex intermediate signal demodulated at the coherent optical receiver is computed at an electric signal processing circuit, and thereby the operation as RC can be performed.

Claims

1. An optical signal processing device comprising: a light source generating an optical signal; first optical modulation means for modulating at least one of intensity and phase of the optical signal at a first modulation period to generate a complex input signal; second optical modulation means for modulating the complex input signal in a time domain at a second modulation period that is shorter than the first modulation period; an optical circulation unit in which the modulated complex input signal circulates at a predetermined delay length; optical multiplex means for joining the modulated complex input signal in the optical circulation unit; a nonlinear response element giving nonlinearity to the optical signal circulating in the optical circulation unit; variable optical modulation means for modulating the optical signal circulating in the optical circulation unit; optical branch means for branching part of the optical signal circulating in the optical circulation unit; optical reception means for demodulating branched light output from the optical branch means to obtain a complex intermediate signal; and a signal processing circuit for weighting each of real and imaginary parts of the complex intermediate signal with any coupling weight and taking a sum to obtain a complex output signal, wherein the signal processing circuit changes the coupling weight so as to reduce an error between the complex output signal and a teacher signal.

2. The optical signal processing device according to claim 1, wherein the modulated complex input signal is a product of a complex vector having a period identical to the first modulation period and the complex input signal.

3. The optical signal processing device according to claim 1, wherein the predetermined delay length is 10 times or more the second modulation period.

4. The optical signal processing device according to claim 1, further comprising optical pulse shaping means for optionally shaping an optical pulse of the optical signal circulating in the optical circulation unit.

5. The optical signal processing device according to claim 4, wherein the optical pulse shaping means comprises: second optical branch means for N-branching (N is an integer of 2 or more) the optical signal circulating in the optical circulation unit; N delay lines being connected to each of N branches of the second optical branch means and having different delay lengths; control means for individually controlling intensity or phase of the optical signal passing through the N delay lines; and optical multiplex means for joining again the optical signal controlled by the control means.

6. The optical signal processing device according to claim 2, wherein the predetermined delay length is 10 times or more the second modulation period.

7. The optical signal processing device according to claim 2, further comprising optical pulse shaping means for optionally shaping an optical pulse of the optical signal circulating in the optical circulation unit.

8. The optical signal processing device according to claim 3, further comprising optical pulse shaping means for optionally shaping an optical pulse of the optical signal circulating in the optical circulation unit.

Description

BRIEF DESCRIPTION OF DRAWINGS

(1) FIG. 1(a) is a view showing a schematic configuration of a typical RC circuit, and FIG. 1(b) is a view showing a schematic configuration of a conventional RC circuit.

(2) FIG. 2 is a view explaining an optical signal processing device according to the first embodiment of the present invention.

(3) FIG. 3 is a view showing a configuration example of a coherent optical receiver.

(4) FIG. 4 is a view showing a schematic configuration of an RC circuit of the present invention.

(5) FIG. 5 is a view showing a configuration of the optical signal processing device according to the second embodiment of the present invention.

(6) FIGS. 6(a) and (b) are views showing configuration examples of an optical pulse shaping unit 521.

(7) FIG. 7 is a view showing a configuration example of a variable optical filter by an optical waveguide formed on a substrate.

(8) FIG. 8(a) is a view of one example showing real parts of a teacher signal and a complex output signal after learning, and FIG. 8(b) is a view of one example showing imaginary parts of a teacher signal and a complex output signal after learning.

DESCRIPTION OF EMBODIMENTS

(9) Hereinafter, embodiments of the present invention will be explained in detail.

First Embodiment

(10) FIG. 2 shows a configuration of an optical signal processing device according to the first embodiment of the present invention. An optical signal processing device 200 of the first embodiment includes an optical modulator 212 controlled by an electric signal processing circuit 210. The optical modulator 212 modulates laser light, which is emitted from a laser light source 211, at a modulation period T.sub.1 either or both of the intensity and phase values of the optical electric field. The complex amplitude of this optical electric field is an input signal u(t). On the other hand, the input signal u(t) is also modulated by the optical modulator 212 at a modulation period T.sub.2 in the time domain so as to be an input signal u′(t).

(11) The converted input signal u′(t) passes through an optical transmission path 213 and enters an optical circulation circuit 215 via an optical coupler 214. The optical circulation unit 215 is loaded with, in addition to the optical coupler 214, a variable attenuator 216, a nonlinear response element 217, and an optical coupler 218. By the optical coupler 218, part of the circulating light is branched into two. One branched light enters the optical coupler 214 via the variable attenuator 216 and circulates in the optical circulation circuit 215. The other branched light is converted into a complex intermediate signal x(t) at a coherent optical receiver 219. This complex intermediate signal x(t) demodulated at the coherent optical receiver 219 is computed by Formula (2) at an electric signal processing circuit 220. Thereby, the operation as RC can be performed.

(12) The input signal u(t) is described by the following formula using a real part term u.sup.r(t) and an imaginary part term u.sup.i(t).
Formula 3
u(t)=u.sup.r(t)+ju.sup.i(t)  (3)

(13) Note that j=(−1).sup.1/2. The signal light u(t) is modulated by some method at the modulation period T.sub.2 (T.sub.2<T.sub.1) in the time domain so as to be u′(t) as in the following formula.
Formula 4
u′(t)=m(t)u(t)  (4)

(14) Note that m(t) is a complex number generated by, for example, the optical modulator. Furthermore, u′(t) may be precomputed in the electric domain to cause the optical modulator 212 to directly modulate u′(t). FIG. 2 realizes the latter method using the electric signal processing circuit 210. The electric signal processing circuit 210 may have an AD conversion function that converts a value computed in the digital domain into an analog value, and in such a case of having the AD conversion function, Formula (4) may be computed in the digital domain. Apart from the modulation period T.sub.2, m(t) has a repetition period T.sub.1 in the following relationship.
Formula 5
m(t)=m(t+T.sub.1)  (5)

(15) Within a range of satisfying the restriction of Formula (5), m(t) can be any value. However, for further excellent learning performance, m(t) is preferred to take a variety of values and, for example, is generated by various pseudo-random number generation algorithms. Furthermore, to prevent divergence of responses, the range which can be taken by m(t) is desired to be restricted to |m(t)|≤1.

(16) Note that, for the optical transmission path 213 and the optical circulation unit 215, for example, optical fibers and optical waveguides can be used. For the optical attenuator 216, a variable attenuator using a Mach-Zehnder interference system or an MEMS mirror can be used to adjust the input light amount. Furthermore, for the nonlinear response element 217, an optical amplifier such as an Er-doped fiber amplifier (EDFA) or a semiconductor optical amplifier (SOA) can be used. The selection of the nonlinear element does not limit the scope of the present invention, which may use, for example, a method that utilizes a laser chaotic oscillation disclosed in Non-Patent Literature 2. Furthermore, in a specific problem, a linear circuit as disclosed in Non-Patent Literature 3 may be configured, without using the nonlinear element 217.

(17) FIG. 3 shows a configuration example of the coherent optical receiver 219. For the coherent optical receiver 219, a typical configuration including a local oscillator light source 310, a 90° hybrid optical circuit 311, and a balanced photodiode 312 can be used.

(18) Learning generalization performance is determined by the diversity of the response of x(t). For securing this diversity, the circulation length T.sub.3 of the circulation unit is desired to be set so as to satisfy the relationship of T.sub.2<<T.sub.3. More specifically, it is desired to be set to T.sub.3≥10T.sub.2.

(19) The complex intermediate signal x(t) obtained at the coherent optical receiver 219 is given as a solution of the following evolution formula.

(20) $\begin{array}{cc}\mathrm{Formula}6& \\ \frac{\mathrm{dx}\left(t\right)}{dt}=\gamma f\left\{\alpha x\left(t-T_3\right)+\beta m\left(t\right)u\left(t\right)\right\}& \left(6\right)\end{array}$

(21) Note that α is the product of the gain of the nonlinear response element 217 and the attenuation amount of the optical attenuator 216, and β and γ are the branch losses of the optical couplers 214 and 218. Here, where T.sub.3=T.sub.1 for simplicity, x(t) is described by a time discretized by the sampling time T.sub.1 as follows.
Formula 7
x.sub.i(n)=f{αx.sub.i(n−1)+m.sub.iu(n−1)}  (7)

(22) Note that n represents the discretized time step. The subscript i means the i-th response of a signal within the sampling time T.sub.1 and further divided by the time T.sub.2. From the relationship described above, i ranges from 1 to N=T.sub.2/T.sub.3. The dynamics of Formula (7), from a comparison with Formula (1), correspond to those of reservoir computing in the case of having a diagonal matrix where all diagonal components of the coupling matrix Ω.sub.ij are jΦ and having the number of neurons being N. That is, the electric signal processing circuit 220 computes Formula (2), and thereby the operation as RC can be performed. Furthermore, the electric signal processing circuit 220 may have an A/D conversion function that converts an analog input into a digital value, and in such a case of having the AD conversion function, computation of signals may be performed in the digital domain. Here, since this configuration handles input and output signals in a complex space, x.sub.i(n), y(n), and ω.sub.i are all complex numbers.

(23) FIG. 4 shows a schematic configuration of an RC circuit of the present invention. As shown in FIG. 4, an RC circuit 400 of the present invention differs from the conventional optical RC (FIG. 1(b)) in that the complex amplitude is the input signal at an input layer 401, and the complex amplitude demodulated at an output layer 403 from the signal output from an intermediate layer 402 can be output as the output signal. This exhibits an excellent function of doubling the effective number of neurons (the sum of the real and imaginary parts is 2N).

Second Embodiment

(24) FIG. 5 shows a configuration of an optical signal processing device according to the second embodiment of the present invention. An optical signal processing device 500 of the second embodiment includes, similarly to the first embodiment, an optical modulator 512 controlled by an electric signal processing circuit 510. The optical modulator 512 modulates laser light, which is emitted from a laser light source 511, at a modulation period T.sub.1 either or both of the intensity and phase values of the optical electric field. The complex amplitude of this optical electric field is an input signal u(t). The signal light u(t) is modulated into an input signal u′(t) at the optical modulator 512, and the converted input signal u′(t) passes through an optical transmission path 513 and enters an optical circulation circuit 515 via an optical coupler 514. The optical circulation unit 515 is loaded with, in addition to the optical coupler 514, a variable attenuator 516, a nonlinear response element 517, an optical coupler 518, and an optical pulse shaper 521. By the optical coupler 518, part of the circulating light is branched into two. One branched light enters the optical coupler 514 via the variable attenuator 516 and circulates in the optical circulation circuit 515. The other branched light is converted into a complex intermediate signal x(t) at a coherent optical receiver 519. The complex intermediate signal x(t) output from the coherent optical receiver 519 is computed by Formula (2) at an electric signal processing circuit 520. Thereby, the operation as RC can be performed. The second embodiment differs from the first embodiment in that the optical circulation unit 515 is provided with the optical pulse shaper 521.

(25) FIGS. 6(a) and (b) show equivalent configuration examples of the optical pulse shaping unit 521. The optical pulse shaping unit 521 includes an M-stage finite impulse response (FIR) filter connected by delay lines of θ, 2θ, . . . , Mθ (θ≤T.sub.1) and uses phase shifters 531 and 531′ and variable attenuators 532 and 532′ to give a weight μ.sub.j to each time component within one pulse of the input signal u(t), that is, each delay line. It is desired to be set to θ=T.sub.1. The weight μ.sub.j of each delay line is a complex number.

(26) When the input signal u(t) modulated at the modulation period T from the optical system as described above is entered into the optical pulse shaper, the optical signal branched by the optical coupler 518 and going to the optical pulse shaper has a time response waveform x(t) described by the following formula.

(27) $\begin{array}{cc}\mathrm{Formula}8& \\ \frac{dx\left(t\right)}{dt}=\gamma f\left\{{.Math.}_{j=1}^M\alpha {\mu }_{j}x\left(t-T_3-\left(M-j\right)T_1\right)+\beta m\left(t\right)u\left(t\right)\right\}& \left(8\right)\end{array}$

(28) Here, μ.sub.j is the weight amount of the j-th (j=1, 2, . . . , M) delay line of the optical pulse shaper 521. M≤T.sub.3/T.sub.1 is desired. Here, where T.sub.3=T.sub.1 for simplicity, consider M≤T.sub.1/2T.sub.3. The following is a case where x(t) is described by a time discretized by the sampling time T.sub.1.
Formula 9
x.sub.i(n)=f{Σ.sub.j=1.sup.MαΩ.sub.ijx.sub.i(n−1)+m.sub.iu(n−1)}  (9)

(29) Here, Ω.sub.ij is as follows.

(30) $\begin{array}{cc}\mathrm{Formula}10& \\ {\left({\Omega }_{\mathrm{ij}}\right)}_{\underset{1\le j\le N}{1\le i\le N}}=\left[\begin{array}{ccccccccccc}{\mu }_1& {\mu }_2& \Lambda & {\mu }_M& 0& & \Lambda & 0& \Lambda & & 0\\ 0& {\mu }_1& {\mu }_2& \Lambda & {\mu }_M& 0& \Lambda & 0& & & \\ 0& 0& {\mu }_1& & \Lambda & {\mu }_M& O& M& & O& M\\ M& M& O& O& O& & O& 0& & O& 0\\ 0& 0& \Lambda & 0& {\mu }_1& {\mu }_2& \Lambda & {\mu }_M& 0& \Lambda & 0\end{array}\right]& \left(10\right)\end{array}$

(31) It can be understood from the symmetry with Formula (1) that this configuration performs the coupling of the intermediate layer in the RC circuit. The number of neurons at this time corresponds to N. Each element of the coupling constant can be set by the weight amount μ.sub.i of each delay line. As compared to the first embodiment, this configuration can set the matrix Ω.sub.ij in a relatively optional manner, which thus has a high capacity to express RC. The operation of the output layer is the same as that in the first embodiment.

(32) A specific implementation method of the FIR filter in the optical domain as described above will be explained. FIG. 7 shows a configuration example of a variable optical filter by an optical waveguide formed on a substrate. In this element, individual ends of a 1:N-branched optical splitter 711 are connected to a delay line group 712 consisting of N delay lines having delay amounts differing by θ, and each delay line is loaded with a variable optical attenuator (VOA) group 713 consisting of N VOAs and a phase shifter group 714 consisting of N phase shifters (see Non-Patent Literature 3). By these elements, the input light is weighted with respect to each time signal and then multiplexed by an optical coupler 715. Thereby, an operation equivalent to that of the FIR filter shown in FIG. 6(b) can be performed.

(33) Although an optical waveguide is used here to form the FIR filter, a spatial optical system can also be used to obtain a configuration equivalent to that in FIG. 6(b). In this case, portions corresponding to the VOA 713 and the phase shifter 714 can be implemented using a spatial light modulator (SLM) or an MEMS mirror.

(34) Learning Method

(35) In RC, a variable to be learned is only ω.sub.i, and several methods are available for determining the variable. As an example, a least mean square (LMS) method described by Formulas (11) and (12) will be explained here, but the present invention is not limited thereto, and the effect of the present invention can be obtained regardless of the algorithm of learning.
Formula 11
ω.sub.i.sup.r(n+1)=ω.sub.i.sup.r(n)+k(d.sup.r(n)−y.sup.r(n))x.sub.i.sup.r(n)  (11)
Formula 12
ω.sub.i.sup.i(n+1)=ω.sub.i.sup.i(n)+k(d.sup.i(n)−y.sup.i(n))x.sub.i.sup.i(n)  (12)

(36) Here, d(n) is a teacher value, and k is a coefficient for determining how much to move in the slope direction. The superscripts r and i indicate the real and imaginary parts for each variable. Since this method merely reduces the energy (error from the learning value) toward the neighboring local minimum, the global search is difficult in this state. Methods for giving an approximation to the global minimum solution include an annealing method. For this too, various methods are proposed. For example, as a function for the time step n, k may be given as follows.
Formula 13
k(n+1)=k.sub.min+h(k(n)−k.sub.min)  (13)

(37) Here, k.sub.min and h are constants.

Learning Example

(38) As a learning example according to the present invention, time series data approximation learning of a complex input and output signal will be shown. NARMA10 task, which is normally used as a benchmark for nonlinear time series learning, is performed to examine whether a teacher signal can be reproduced. The optical system of the optical signal processing device according to the first embodiment of the present invention is reproduced in the simulation to compute whether the output of NARMA10 described by Formula (14) can be approximated.

(39) $\begin{array}{cc}\mathrm{Formula}14& \\ y\left(n+1\right)=0.3y\left(n\right)+0.05y\left(n\right)\underset{i=0}{\overset{9}{.Math.}}y\left(n-i\right)+1.5u\left(n\right)u\left(n\right)+0.1& \left(14\right)\end{array}$

(40) Here, y(n) is a time series signal to be predicted, and u(n) is an input signal. For the nonlinear element, the input signal u(n) is generated by Formula (15) below.

(41) $\begin{array}{cc}\mathrm{Formula}15& \\ u\left(n\right)=\sin \left(2\pi f_1\frac{t}{T}\right)\sin \left(2\pi f_2\frac{t}{T}\right)\sin \left(2\pi f_3\frac{t}{T}\right)& \left(15\right)\end{array}$

(42) Here, f.sub.1, f.sub.2, and f.sub.3 are 2.11, 3.73, and 4.33, respectively. The modulation period T.sub.2 of the mask function m(t) is set to T.sub.2=T.sub.1/100, and the circulation time T.sub.3 is set to T.sub.3=4T.sub.1=400T.sub.2. For the nonlinear element, an SOA is used and its nonlinear dynamics are computed by a method disclosed in Non-Patent Literature 4. The initial values of the weight vector ω.sub.i in the output layer to be learned are all set to 1. Furthermore, α, which is the constant for determining the mutual coupling matrix in the intermediate layer of the network, is selected to be 1.2. As α increases, the dynamics that constitute the reservoir become chaotic. Accordingly, α=1.2 is set so as to maximize the reservoir network within a range of showing no chaotic property. Setting in this manner increases storage capacity of the reservoir network, exhibiting an excellent function of improving learning performance for tasks including past information as in NARMA.

(43) The learning is performed using an LSM method. A teacher signal of 1000 symbols is learned and then 1000 symbols are estimated. FIGS. 8(a) and (b) show one examples of the real and imaginary parts of the teacher signal and the complex output signal after learning. Here, the learning is performed under the condition where the number of nodes N=100. As shown in FIGS. 8(a) and (b), the waveform of the complex output signal after learning gives a good approximation to the waveform of the teacher signal, and the normalized mean square error (NMSE) described by Formula (15) is 0.01, which is sufficiently small. Therefore, by using the configuration of the present invention, learning of a complex signal can be performed.

REFERENCE SIGNS LIST

(44) 100, 400 RC circuit 101, 401 Input layer 102, 402 Intermediate layer 103, 403 Output layer 200, 500 Optical signal processing device 210, 220, 510, 520 Electric signal processing circuit 211, 511 Laser light source 212, 512 Optical modulator 213, 513 Optical transmission path 214, 218, 514, 518 Optical coupler 215, 515 Optical circulation unit 216, 516 Variable attenuator 217, 517 Nonlinear response element 219, 519 Coherent optical receiver 521 Optical pulse shaper