JP4802329B2 - Eye position measuring method and eye position measuring apparatus - Google Patents

Description

  The present invention relates to a method and apparatus for measuring an eyeball position, that is, an eyeball position, and in particular, an eyeball position measuring method and an eyeball position measuring apparatus for estimating the eyeball position by measuring the distribution of the peripheral electric potential of the eye using three or more electrodes. About.

The human eyeball is operated as an ion pump in the retinal neurons, so that current flows from the retinal side to the corneal side, so the corneal side is considered as a battery model with a positive pole (+) and a retinal side with a negative pole (-). It is done. That is, the eyeball behaves as an electric dipole having a positive charge on the cornea side and a negative charge on the retina side. Therefore, when the eyeball is moved left and right and up and down, the potential distribution around the eyeball changes accordingly. For example, when the eye is directed to the right side, the potential around the right side of the eye is higher than the potential around the left side, and therefore the right side is higher when the skin surface potentials on the right and left sides of the eye are compared.
As described above, a method for estimating the position of the eyeball in the horizontal direction or the vertical direction using the potential difference between the pair of electrodes attached to the left and right or the top and bottom of the eye by using the fact that the eyeball is charged is particularly medical. In the clinical field, as an electrooculogram (EOG) technique, it is often used as a means mainly for examining characteristics of nystagmus.

  Non-patent document 1 is a paper in which the potential predicted from a model in which the eyeball is regarded as a battery is compared with the potential obtained from an observed electrooculogram and the effectiveness of the battery model is verified. In this paper, the electric potential in the vicinity of the outside of the eyeball is measured when the eyeball is subjected to an impulse motion alternately left and right around its front. Then, the change in potential when the eyeball is moved is derived as an EOG potential, and the potential change due to the difference in the position around the eyeball is analyzed.

There is also proposed a device that measures the effective visual field range of the eyeball and measures the potential of the eyeball by attaching electrodes one by one to the left, right, and top of the eyeball. (For example, refer to Patent Document 1).
The eye potential measurement method described in Patent Document 1 measures each eyeball position change in the horizontal direction and vertical direction of the eyeball as a potential difference, and extracts a high-speed eye movement component from the horizontal and vertical change output. is doing. Then, the effective visual field range is measured by measuring the number of times of the high-speed eye movement, performing a count comparison for each visual field position, and obtaining the position of the target stimulus that can be detected by the high-speed eye movement once or less. .

Izuki, Kubo, Shiraishi, Nishikawa, Mimura “Battery model of eyeball for eyeball standing potential analysis” The Japanese Society of Ophthalmology 99 9 9 1012-1016 Japanese Patent Laid-Open No. 9-38037

  In general, the EOG technique uses a correlation between a potential difference between a pair of electrodes and an eyeball position (a linear relationship is established in a wide range), and potential differences corresponding to several eyeball positions in advance. Is used to obtain a linear regression equation (calibration), and the eye position is estimated from the potential difference using this regression equation. Although this EOG method is recognized for its simplicity and usefulness, it can be fully used as a method for measuring eyeball position due to measurement accuracy problems, measurement environment / conditions, and unstable factors such as individual differences. It has not yet reached.

The reason for this is that the regression equation indicating this linear relationship changes the slope of the linear relationship over time due to the drift phenomenon in which the current density (or charge) of the eyeball changes over time, so that the position of the eyeball stable over the long term It is in the point that cannot be estimated. The cause of the drift phenomenon is due to a change in the function of the ion pump of the retinal nerve cell, and its magnitude is thought to depend on the intensity of light entering the eye, but it is impossible to determine the change in advance. As a result, it has been difficult to compensate for the drift phenomenon.
Therefore, as an eyeball position measurement method, for example, a video camera or the like is arranged at a position where the eye movement can be observed, and thereby, the eye movement itself or the movement of the virtual image of the light source transferred to the eye is observed. A method such as an image analysis method or a scleral reflection method in which light is applied to the eye and the intensity change of the reflected light is observed is employed.
However, when such a method is used, the optical system or video camera installed in front of the measurement subject's eyes obstructs the visual field, and the head needs to be equipped with an optical system, a signal processing circuit, etc. The physical and psychological burden on the measurement subject is large. Normally, these devices are connected to an external main device through a cable to perform signal processing, but at this time, they may be affected by noise during signal transmission or restrict the movement of the measurement subject. There are problems such as. Therefore, these methods are difficult to use because measurements outside the laboratory, such as long-time measurements, environments where the measurement subject's body moves, and measurements at bedtime, interfere with the head-mounted device. There was a problem.

  The EOG method described in Non-Patent Document 1 is based on a conventional data analysis method based on correlationism, and a correlation between a potential difference between a pair of electrodes and an eyeball position (a linear relationship is established in a wide range). Rather than model-based thinking, that is, the eyeball is regarded as a battery model, and the electrooculogram is analyzed based on the electrode position, the eyeball size, the eyeball position, and the causal relationship between the current density and the skin potential around the eye. The conventional EOG method is improved in a systematic form. However, this method also has the same problem as described above that compensation for the drift phenomenon is not made.

  In addition, the device described in Patent Document 1 obtains an effective visual field range in which a target stimulus can be detected on the visual field position. In this example, the potential measurement is performed by attaching electrodes to the left and right or upper and lower sides of the eyeball. In a certain point, like the EOG method described above, there is a problem that stable measurement cannot be performed due to a drift phenomenon in which the potential changes with time.

  The present invention effectively suppresses the influence of the drift phenomenon by dynamically estimating the magnitude of the current density in the battery model, with high accuracy, and without further restricting the movement of the measurement subject's body. It is another object of the present invention to provide an eye movement measurement method and an eyeball measurement apparatus which can be performed without obstructing the visual field.

  In order to solve the above-mentioned problems and achieve the object of the present invention, the eyeball position measuring method of the present invention combines the potentials detected by at least three or more electrodes attached around the eye (eyes) as appropriate. The difference voltage is taken out, and the current density flowing from the retina side to the cornea side of the eyeball or the retina and the cornea are considered to be charged in a model in which the eyeball is regarded as a battery using information on the plurality of taken out potential differences. Estimate the size of the electric charge, the position of the eyeball, the position where the electrode is applied, and the size of the eyeball (eyeball radius), and then use the estimated electrode application position and the eyeball radius to extract the potential difference. The eye position and the current density are estimated using an EM (Expectation and Maximization) algorithm method that alternately repeats the estimation of the eye position and the current density over a plurality of time cycles. The present invention is characterized in that the current density or the magnitude of electric charge that is considered to be charged in the retina and cornea is estimated.

  An eyeball position measuring apparatus according to the present invention includes at least three or more electrodes mounted around the eye, a signal conversion circuit that detects a potential difference between the plurality of electrodes from an arbitrary combination between the electrodes, From the potential difference extracted from the signal conversion circuit, the current density that flows from the retina side to the cornea side of the eyeball or the magnitude of the charge that is considered to be charged to the retina and cornea in the model where the electrode attachment position and the eyeball are regarded as batteries A processor and a data processing circuit for estimating the position by repetitive calculation. In the calculation of the data processing circuit, the estimation of the eye position and the estimation of the current density are performed in a plurality of cycles by an EM (Expectation and Maximization) algorithm method. It is characterized in that it is calculated over a range.

  The eyeball measuring device according to a preferred embodiment of the present invention includes a transmission circuit that converts a potential difference obtained from the measured potential into a digital signal and transmits the signal, and a signal transmitted from the transmission circuit by wire or wirelessly. And a receiving circuit for receiving.

According to the eyeball measurement device and the eyeball movement measurement method of the present invention, the eyeball position (eye orientation) and current density are estimated based on the potential signals obtained from three or more electrodes attached to the periphery of the eye. Since this is repeated a predetermined number of times for a predetermined time using the EM algorithm, it is possible to follow a temporal change in current density, and as a result, it is possible to estimate an accurate eyeball position. It has become. Originally, the EOG method can measure by simply attaching an electrode around the eyeball, so there is no need to block the visual field of the person to be measured or mount an optical system on the head.
In addition, in a preferable eyeball measuring device, the observed potential signal can be sent to a remote receiving device using digital radio or the like, and data processing can be performed. It is possible to measure without being limited by the time.

According to the eyeball position measuring method and the eyeball position measuring apparatus of the present invention, the eyeball position (eye orientation) and current density are estimated based on potential signals obtained from three or more electrodes attached to the periphery of the eye. Since this is repeated a predetermined number of times for a predetermined time using the EM algorithm, it becomes possible to follow the temporal change in the current density, and as a result, an accurate eyeball position can be estimated. It is possible.
In addition, since the EOG method can measure by simply attaching an electrode around the eyeball, it is not necessary to block the visual field of the person being measured or to mount an optical system on the head. Since data processing can be performed by sending data to a remote receiving device using digital radio or the like, measurement can be performed regardless of where the measurement target person is and without being limited by the measurement time.

  In addition, the eye movement measuring device according to a preferred embodiment of the present invention is capable of measuring from a remote location, and converts a potential difference obtained from the measured potential into a digital signal and transmits the digital signal. And a receiving circuit for receiving a signal transmitted by wire or wireless.

  In addition, according to the present invention, the eyeball position can be measured simply by attaching an electrode around the eye or the like, so that restrictions imposed on the measurement target can be minimized, and there is a problem in the EOG method. It is possible to suppress the influence caused by the drift phenomenon.

  In the present invention, since the EM algorithm is used, in addition to the eyeball current density or the eyeball charge amount, it is also possible to estimate on-line unknown parameters such as the electrode sticking position. Calibration work can be eliminated.

  The present invention relates to an eyeball position measuring method and an eyeball position measuring apparatus for detecting a rotation position of an eyeball, that is, a pupil position, by a change in skin potential around the eye that occurs when a human eyeball is regarded as a battery model and the eyeball is rotated. However, first, the battery model of the eyeball will be described prior to the description of the embodiment of the present invention.

FIG. 1 is a battery model diagram of an eyeball.
Using the human eyeball 21 as a spherical model, the center of eyeball rotation is O, the center of the cornea is A, the center of the retina is B, the electrode position on the skin around the eye is M, and the cornea center position A is from the electrode position M. Is r, and the distance from the electrode position M to the retina center position B is r ′.
As shown in FIG. 1, it is assumed that the current I flows from the retina to the cornea side in the eyeball. At this time, the potential V _M of the point M can be expressed by the following equation (1) below. Here, it is assumed that the eyeball is seen in a surrounding tissue having a uniform electrical conductivity σ.

  FIG. 2 is an explanatory diagram for setting each axis, each angle, and the like. As shown in FIG. 2, when the rotation center of the eyeball 21 having a radius a, that is, the rotation center of the pupil is O, the axis that penetrates from the back of the head to the face is x, the horizontal axis orthogonal to this axis x is y, and the vertical axis is z. Further, the angle at which the eyeball rotates with respect to the z-axis is defined as a horizontal angle θdeg, and the angle at which the eyeball rotates with respect to the y-axis is defined as a vertical angle ξdeg. At this time, assuming that the coordinates of the point M are x = (x, y, z) and the direction of the eyeball is (θ, ξ), the distances r and r ′ can be expressed by the following (Equation 2).

  Further, when the current density (I / 4πσ) is simply I, the above (Equation 1) can be expressed as (Equation 3).

ここで、ｒ、ｒ’が、電極貼付位置ｘ=(ｘ、ｙ、ｚ)と眼球半径ａ及び眼球位置θ、ξの関数であることから、この電位は、電極位置、眼球半径、電流密度と眼球位置の関数であることがわかる。

Here, since r and r ′ are functions of the electrode attachment position x = (x, y, z), the eyeball radius a, and the eyeball positions θ and ξ, this potential is the electrode position, eyeball radius, and current density. And a function of the eyeball position.

Here, in practice, the current density I is changed from moment to moment, it occurs drift phenomenon in which the potential V _M varies.
In the present invention, the eyeball position is estimated by an EM (Expectation and Maximization) algorithm using the battery model described above.

  In the present invention, the objective is to estimate the eyeball position. To this end, it is necessary to estimate the electrode position, eyeball radius, and current density prior to this. However, since neither of these can be obtained directly, these unknowns are estimated simultaneously by the method described below.

  The estimation method differs depending on whether the number of unknowns is one or two. When estimating one kind of unknown, the unknown is estimated by searching for an unknown that minimizes the square error of the electrooculogram based on the measured electrooculogram obtained from the multipoint electrode and the battery model of the eyeball (least square estimation). ). On the other hand, when estimating two kinds of unknowns, 2 based on the principle of the EM algorithm using the electrooculogram obtained from the multipoint electrode and the electrooculogram based on the battery model of the eyeball, that is, by the iterative calculation that repeats the least squares method. Try to obtain an estimate of the kind unknown at the same time.

In the model of the present invention, the estimated values are classified into eyeball position, current density, electrode position, and eyeball radius. Of these, the latter electrode position and eyeball radius are constants that hardly change once measurement is started, and therefore need only be estimated once at the start of measurement (however, this denies re-estimation during measurement). Not a thing). On the other hand, since both of the former eyeball position and current density change with time, it is necessary to repeat estimation every moment.
Among the former, the eyeball position is a variable that changes in milliseconds (especially, when a high-speed eyeball position displacement of eye movement called saccade occurs, the maximum speed may exceed 500 deg to 600 deg / sec. For example, refer to Patent Document 1.) Since the time change of the current density occurs on a time scale on the order of seconds (this is considered that the change of the ion pump of the retina appears macroscopically), both estimation methods are Different. That is, the eye position needs to be updated on the order of milliseconds to 10 milliseconds, while the current density should be updated on the order of seconds to 10 seconds. Therefore, when the eyeball position is actually measured, the current density is estimated based on a plurality of data included in a time interval in which the value can be regarded as almost constant, and the eyeball position in that interval is used by using the estimated current density. Is estimated.

First, a method for estimating the electrode attachment position and the eyeball radius will be described. Although these unknowns can be estimated online without calibration work, here, a method for performing calibration using an index will be described. 2 and 10B are explanatory diagrams of the setting conditions.
Now, the attachment positions of the M electrodes are assumed to be X _m = (x _m , y _m , z _m ) (m = 1,..., M), respectively. Then, a value obtained by obtaining N potential differences between the electrodes of the multipoint electrode for each of the K reference points, ΔV _nk (n = 1, 2,..., N, k = 1, 2,..., K) and the battery of the eyeball A value that minimizes the square error of ΔV _nk ′ calculated by the model is obtained. The potential difference calculated from the battery model is a function of the electrode position, the eyeball radius, the current density, and the eyeball position. Among these, the eyeball position can be calculated from the position of the reference point that the measurement subject gazes at. Therefore, the remaining electrode position, eyeball radius, and current density are unknown. Therefore, based on the idea of least square estimation, E ({X _m }, a, I) in the following (Equation 4) is determined to be minimized ({X _m }, a, I), and this is calculated as the electrode position, eyeball. Estimate radius and current density.

なお、この推定を行なうためには、複数の参照点を注視することが必要である。また、複数の電極があらかじめ一つのパッケージ上に装着されていてそれらの相互位置が限定されている場合には、電極位置を表す変数を個々の電極に対して独立して定める必要はなく、電極の構造に応じて推定すべき未知数を適切に定めればよい。

In order to perform this estimation, it is necessary to watch a plurality of reference points. In addition, when a plurality of electrodes are mounted on a single package in advance and their mutual positions are limited, it is not necessary to independently define a variable representing the electrode position for each electrode. What is necessary is just to determine appropriately the unknown which should be estimated according to the structure.

  Next, a method for simultaneously estimating the eyeball position and the current density will be described. As described above, the value of the current density changes with time in the order of seconds, but the order of the change is sufficiently longer than the order of changes in the eyeball position (eye movement). The eyeball position estimation needs to be performed in a cycle of milliseconds, but the current density can be considered to be constant in a certain time interval in which this estimation is performed. In the present apparatus, this time interval is set to about 1 to several tens of seconds, and the current density is estimated for each interval and used for eyeball position estimation within that interval. Then, an EM algorithm is used to obtain these two kinds of unknowns simultaneously.

It is assumed that the electrode position and the eyeball radius have already been correctly estimated by the previous operation.
And the initial value of the current density used for repeated calculation is determined appropriately, and this is expressed as I ^BUF .
The potential difference ΔV _n ′ based on the battery model of the eyeball is a function of the electrode position, the eyeball radius, the current density and the eyeball position, but the values of the electrode position and the eyeball radius are already known, and the current density value If the initial value determined in is used, the potential difference is a function of only the eyeball position. Therefore, by searching for the eyeball positions θ and ξ that minimize the square error between the potential difference ΔV _n (n = 1, 2,..., N) observed from the electrodes and ΔV _n ′ calculated by the battery model of the eyeball. An estimated value of the eyeball position can be obtained (Equation 5). That is, if there are N _i measurement data in the measurement section, the same is repeated for each data, so that N _i sets of eyeball position estimation values (θ _i , ξ _i ) (i = 1, 2). , ..., N _i ).

しかし、このようにして得られたＮ_ｉ組の眼球位置推定値は、適当に定めた電流密度の初期値Ｉ^ＢＵＦに依存したものであって、一般に真の眼球位置と一致しない。したがって、正しい眼球位置推定値を得るためには、正しい電流密度の値を知らなければならない。
ＥＭアルゴリズムでは、電流密度の値を推定するために、先のステップで仮に推定されたＮ_ｉ個の眼球位置推定値を利用する。
繰り返しになるが、眼球の電池モデルの基づく電位差ΔＶ_ｎ´は、電極位置、眼球半径、電流密度と眼球位置の関数であり、電極位置と眼球半径の値はすでにわかっている。ここで、前のステップで得られた眼球位置推定値（θ_ｉ、ξ_ｉ）として、これを正しい眼球位置であるとみなせば、電位差ΔＶ_ｎ´は電流密度だけの関数になる。そこで、今度は、電流密度が未知数であると考えて、電極から観測された電位差ΔＶ_ｎｉ（ｎ＝１、２…、Ｎ、ｉ＝１、２、…、Ｎ_ｉ）と眼球の電池モデルで計算したΔＶ_ｎｉ’の２乗誤差が最小になるような電流密度Ｉを求め、これを電流密度の推定値とする（数６）。このとき、Ｎ_ｉ個の観測区間では電流密度は一定であると仮定しているので、誤差の計算には、Ｎ_ｉ個の観測データΔＶ_ｎｉと眼球位置推定値（θ_ｉ、ξ_ｉ）をすべて利用する。

However, the N _i sets of estimated eyeball position values obtained in this manner depend on an appropriately defined initial value I ^BUF of the current density and generally do not coincide with the true eyeball position. Therefore, in order to obtain a correct eyeball position estimation value, the correct current density value must be known.
The EM algorithm, in order to estimate the value of current density, using the tentatively estimated N _i number of eye position estimate in the previous step.
Again, the potential difference ΔV _n ′ based on the battery model of the eyeball is a function of the electrode position, eyeball radius, current density and eyeball position, and the values of the electrode position and eyeball radius are already known. Here, if this eyeball position estimated value (θ _i , ξ _i ) obtained in the previous step is regarded as the correct eyeball position, the potential difference ΔV _n ′ is a function of only the current density. Therefore, this time, assuming that the current density is unknown, the potential difference ΔV _ni (n = 1, 2,..., N, i = 1, 2,..., _Ni ) observed from the electrodes and the battery model of the eyeball A current density I that minimizes the square error of the calculated ΔV _ni ′ is obtained, and this is used as an estimated value of the current density (Equation 6). At this time, since the N _i number of observation interval is assumed to be the current density is constant, the calculation of the error, N _i number of observed data [Delta] V _ni and the eyeball position estimate (θ _i, ξ _i) the Use all.

このようにして新しい電流密度の推定値Ｉを得ることができた。次のステップでは、こうして得られた推定値を再び電流密度の初期値Ｉ^ＢＵＦとしてみなして、上記の計算を繰り返す。そして、電流密度の推定値や眼球位置の推定値の繰り返しによる変動が十分に小さくなった段階で繰り返し計算を打ち切り、その時点で得られた眼球位置推定値を最終的な眼球位置推定値として出力する。
このような繰り返し計算の結果得られる推定値が常に真の値と一致することは理論的には保証されていないが、十分な数のデータと真の値に近い初期値から計算を始めれば、多くの場合十分によい精度で未知数を推定することができる。特に、本装置の場合は、未知数である電流密度のおおよその値は経験的にわかっていることから、繰り返し計算は十分によい初期値から出発することができる。また、いったん計測がはじまれば、連続する計測区間のあいだで電流密度の変動は小さいと考えられるので、前計測区間で得られた電流推定値を当該計測区間の電流推定値の初期値として用いることにより、少ない繰り返し回数で正しい推定値が得られるものと期待される。

In this way, a new current density estimate I could be obtained. In the next step, the estimated value thus obtained is regarded as the initial value I ^BUF of the current density again, and the above calculation is repeated. Then, the calculation is aborted when the fluctuation due to the repetition of the estimated current density value and the estimated value of the eyeball position becomes sufficiently small, and the eyeball position estimated value obtained at that time is output as the final estimated eyeball position value. To do.
Although it is not theoretically guaranteed that the estimated value obtained as a result of such repeated calculation always matches the true value, if the calculation is started from a sufficient number of data and an initial value close to the true value, In many cases, unknowns can be estimated with sufficiently good accuracy. In particular, in the case of this apparatus, since the approximate value of the current density, which is an unknown, is known empirically, iterative calculations can start from a sufficiently good initial value. In addition, once the measurement starts, the current density fluctuation is considered to be small between consecutive measurement intervals, so the current estimation value obtained in the previous measurement interval is used as the initial value of the current estimation value in the measurement interval. Therefore, it is expected that a correct estimated value can be obtained with a small number of iterations.

  Numerical values are used to investigate whether or not eyeball position estimation by the above algorithm is possible in principle, how much error can be estimated, and what electrode arrangement is suitable for eyeball position estimation. The experiment was conducted.

In this numerical experiment, the radius of the eyeball was set to 12 mm, and for simplicity, this value was already estimated. Further, as schematically shown in FIG. 3 (1) to (6), the number of electrodes 21 arranged around the eyes (eyes) 23 is three, four, and five, and the electrodes 22. As the arrangement conditions, two patterns were prepared: horizontally arranged under the eyes 23 and L-shaped. In addition, in order to simulate electrical noise and measurement error applied to the electrode, the potential value measured at the electrode is 0 mV on average calculated from the eyeball battery model, with a standard deviation of 0.01 to 0.00. A value obtained by adding Gaussian noise of 1 mV (dispersion 0.0001 to 0.01 mV ² ) was used.

  In the first experiment, the eyeball position was first estimated under the condition that all variables other than the eyeball position were known. In the experiment, the horizontal eyeball position (θ) is set every 2 deg from −20 deg to 20 deg, and the vertical eyeball position (ξ) is set every 0.5 deg from −5 deg to 5 deg. Asked.

FIG. 4 is a diagram showing average estimation errors for the six types of electrode arrangements in FIG. As can be seen from this result, the position estimation error in both the horizontal and vertical directions varies slightly depending on the electrode arrangement, but when the standard deviation of noise is about 0.03 mV (variance is 0.001 mV), the horizontal and vertical The eyeball position can be estimated with an estimation error of about 0.1 deg in any direction.
As described above, it was found that the eyeball position can be estimated in principle.

  Next, an experiment was performed to estimate the current density and the electrode position at the same time under the condition where the eyeball position is known (corresponding to the work at the start of measurement).

Here, assuming that the true value of the current density is I = 25 Am / S, the line-of-sight directions (θ, ξ) are (−15, 0), (0, 0), (0, 15), (− 5, 0) and (5, 0) were used to measure the potential, and the current density and the electrode position were estimated simultaneously. FIG. 5 is a diagram showing the estimation error of the result.
As can be seen from this figure, under the condition that the standard deviation of noise is about 0.03 mV, when the number of electrodes is 4, the x, y, and z coordinates can be estimated with an estimation error of about 0.1 mm. On the other hand, the current density can be estimated with an error of about 0.2 Am / S.
As described above, it has been shown in principle that the electrode position and the current density can be estimated under conditions where the eyeball position is known.

Finally, an experiment was performed to estimate the eyeball position and current density at the same time under conditions where the electrode position was known. Also in this experiment, the true value of the current density is I = 25 Am / S, and the eyeball positions are (θ, ξ) = (− 15, 0), (0, 0), (0, 15), (−5). , 0) and (5, 0) were taken at five locations. The initial value of the current density was obtained when I = 20 Am / S and I = 30 Am / S.
FIG. 6 is a graph showing changes in the estimated value of the current density in the repeated calculation. As can be seen from this graph, it can be seen that the estimated value of the current density changes toward the true value from any initial value after several iterations. In this calculation, when the estimated value of the eyeball position is obtained, since it is set to a discrete value having a step size of 0.01 deg of the eyeball position, the current density value does not change during the repeated calculation and becomes a true value. Although not reached, it is considered that a better estimated value can be obtained by estimating the estimated eyeball position as a continuous value as a variable.
As described above, it has been shown in principle that the eyeball position and the current density can be estimated under the condition where the electrode position is known.

Next, an embodiment of the present invention will be described, but the present invention is not limited to this embodiment.
First, an example of an embodiment of an eyeball position measuring device that implements the eyeball position measuring method according to the present invention will be described.
FIG. 7 is a block diagram showing the relationship of hardware components necessary to implement the present invention. In practicing the present invention, three or more electrodes are required, but in the example shown in FIG. 7, five electrodes 2 (2a to 2e) are used. The electrodes 2a to 2e are attached in the vicinity of one eye (eye) 1 of the measurement subject, for example, in the lower part or the upper part.
The electrodes 2 a to 2 e are connected to a signal conversion circuit 3, and the signal conversion circuit 3 is connected to a data processing circuit 4.

  The overall calculation and control of the embodiment of the present invention is performed by a central processing unit (CPU) 5. The CPU 5 includes a RAM 7 that stores signals detected by the electrodes 2a to 2e, and a ROM 8 that stores various programs for executing the above-described EM algorithm. In addition, a timer 6 for measuring time for performing the measurement of the present invention in a predetermined cycle, and a display device 9 for displaying a change in eyeball potential and an estimated current are provided. The measurer looks at the display device 9 and can estimate the eyeball position of the measurement subject.

  In the block diagram shown in FIG. 7, first, when the eyeball 1 moves left and right or up and down, the potentials detected by the electrodes 2a to 2e change variously. The potential signals obtained from the respective electrodes 2a to 2e are sent to the signal conversion circuit 3, where the potential difference (voltage) between the electrodes is taken out. These potential differences are processed in the data processing circuit 4 under the control of the CPU 5 as will be described later, and the results are displayed on the display device 9. The processed data is stored in the RAM 7.

FIG. 8 is a block diagram showing another embodiment of the present invention. The same components as those in FIG. 7 are denoted by the same reference numerals, and description thereof is omitted. As shown in FIG. 8, in this example, the voltage signal between the electrodes from the signal conversion circuit 3 is sent to the transmission circuit 11, and is sent to the reception circuit 12 through transmission means such as radio. Although not shown in FIG. 8, the receiving circuit is coupled to the computer system of FIG. 7, and similar data processing is performed in the data processing circuit 4 under the control of the CPU 5 (see FIG. 7).

Next, an embodiment of an eyeball position, that is, an eyeball position measuring method according to the present invention will be described.
[First Embodiment of Eyeball Position Measurement Method]
This measurement is performed by the EM algorithm. FIG. 9 is a flowchart of this embodiment, FIG. 10 (a) is a flowchart in the calibration step A for obtaining the initial value, and FIG. 9 (b) is an explanatory diagram for that. FIG. 11 is a time chart schematically showing a series of steps for measuring the eyeball position (eyeball position). For the measurement of the eyeball position, as described above, a plurality of electrodes are attached around the eye of the measurement target person (five examples are shown in FIGS. 7 and 8), etc. It is made by wearing.

The measurement of the eyeball position is performed in a plurality of measurement sections E (E1, E2, E3...) As shown in the time chart of FIG. 11, and the current density described above is substantially constant in these sections. For example, about 1 second to several tens of seconds of the time length that can be recognized to be selected. The eyeball position changes even within this time.
Further, it is assumed that the mounted electrode position X and eyeball radius a do not change with respect to the entire measurement section E.

The operation of the first embodiment of the present invention will be described with reference to the flowchart of FIG.
First, as an initialization operation (calibration), an electrode position and an eyeball radius are estimated (step S1), and an initial value I ^BUF of the current density is set (step S2). Details of the calibration process will be described with reference to FIG.
As shown in FIGS. 10A and 10B, first, as described above, three or more electrodes are mounted around the measurement subject (step S20).
On the other hand, the reference surface 30 to the eye front of the measured person by directly facing coordinate plane (and x-y coordinate plane) provided perpendicular to the reference plane 30, and the axis z passing through the eye center e _O-measured person , x, the origin of the intersection of the y both axes, the reference point _{p k} coordinate points of the plurality of indicating with a × mark on the reference surface 30 (K _{pieces) (r xk, r yk)} (k is the reference point No.) (Step S21).
Then, the angle θ _k of the x-axis direction component and the angle ξ _k of the y-direction component with respect to the z-axis passing through the origin of the reference surface 30 of the eyeball position when these reference points _pk are watched are expressed by the following (Equation 7). (Step S22).

（ここで、Ｄは眼球中心ｅＯから特定された参照面３０までの距離である。）

(Here, D is the distance from the eyeball center eO to the identified reference plane 30.)

Then, sequentially to gaze at the K reference points _{p k} to measured person (Step S23), the potential difference from each electrode to each reference point _{p k ΔV nk (n = 1,2} ..., N, k = 1,2 ..., K) are acquired (step S24). In order to keep the current density in the eyeball during the measurement operation constant, consideration is given so that the condition of light incident on the eye of the measurement subject is kept constant.
Subsequently, under the assumption that the electrode position, eyeball radius, and current density are constant during the K measurements, the evaluation function E ({X}, a, I) (number The values of the electrode position x, the eyeball radius a, and the current density I that minimize the value of 4), that is, argmin E ({X}, a, I) are obtained (step S25).
In this way, the electrode positions and eyeball radii that can be considered constant during measurement are estimated. Moreover, the initial value of the current density used in the subsequent estimation is also obtained at the same time (step S26).
The above is the calibration process in FIG. 9, that is, the setting of the electrode position, eyeball radius estimation (step S1) and the current density estimation value ^IBUF as the current initial value. (Step S2).

Returning to FIG. 9 again, next, an initial reset, that is, a timer 6 on which an operation is performed is reset. (Step S3). That is, the time T of the timer 6 is set to “0”.
And the eyeball position measurement of the 1st measurement area E1 shown in FIG. 11 is performed. In this eyeball position measurement, the time T set in step S3 is set to the timer time t, and a counter (not shown) is set to an initial value 1 (step S4). Then, to get the potential difference [Delta] V _n from each electrode attached to the measured person at time t, put to as [Delta] V _ni (step S5). Using the obtained potential difference ΔV _n and the estimated current density I ^BUF obtained by the calibration in the step A, the least square estimated values θ, ξ that minimize the evaluation function E (θ, ξ; I ^BUF ), that is, argminE (θ, ξ; I ^BUF ) is obtained (Equation 5) (step S6). Thus, the eyeball position (θ _i , ξ _i ) at time t (counter value _i ) is estimated. Further, the obtained θ _i and ξ _i are stored as θ ^BUF _i and ξ ^BUF _{i in} the i-th buffer (step S7). These steps are referred to as process B.

Next, the time t is allowed to warm to t + Delta] t (step S8), and the counter is incremented (substitutes i + 1 obtained by adding 1 to i) (step S9), and by measuring the number of times i has reached the required number _{N i} It is determined whether or not (determination step S10). The steps from step S5 to step S10 are repeated until the number of times of measurement _i reaches Ni.
When the measurement number i becomes i ≧ _{N i,} the estimated value of _{N i} number which has been stored in the buffer ^{_{θ BUF i, ξ BUF i (}} i = 1, ... N i) using a least square method The current density I that minimizes the evaluation function E (I; {θ ^BUF _i }, {ξ ^BUF _i }), that is, argminE (I; {θ ^BUF _i }, {ξ ^BUF _i }) is calculated (Equation 6) (Step S11). The estimated value I of the current density obtained by this calculation is stored in the buffer as ^IBUF (step 12). This process is referred to as process C.

The steps B and C are schematically shown by broken lines in FIG. While i <N _i described above, the steps S5 to S10 are repeated in the section F1 shown in FIG.

Then, the eyeball position theta in the counter value i, and determines the estimated value theta ^BUF _i of xi], and xi] ^BUF _i, whether the estimated current density ^{I BUF} has converged (step S13), and it is determined that they have converged Step D from Step S4 to Step S13 is continued for the sections F2, F3... Shown in FIG. If it is determined in step S13 that θ ^BUF _i , ξ ^BUF _i , and I ^BUF have converged, the measurement of the eyeball position in the section E1 in FIG. 11 is terminated, and θ ^BUF remaining in the buffer at that time _i and ξ ^BUF _i are output as final estimated values of the eyeball positions.
Then, the start time T + ΔT is set (step 14), the process returns to step S4, and measurement of a new ΔT section (section E2 in FIG. 11) is started. Note that the initial value of the current density is not the initial value obtained by the calibration described above, but the I ^BUF obtained in the section E1 preceding the section E2. In the situation where the temporal change of the current density is small, it is expected that the number of iterations until the iteration calculation converges is greatly reduced by using the estimated value obtained in the previous section as the initial value. In this way, the measurement of E3, E4... Is executed sequentially, and the estimated amount of the eyeball position is obtained.

In the first embodiment of the eyeball position measurement according to the present invention described above, as shown in FIG. 11, in the sections E2, E3,... After the section E1, the estimated amount θ converged in the preceding section. ^{Although BUF} _i , ξ ^BUF _i , and I ^BUF are used as initial values, the eyeball position can be measured by an irregular EM algorithm different from the above. An embodiment in this case will be described.

[Second Embodiment of Eyeball Position Measuring Method]
FIG. 12 is a flowchart showing an anomalous EM algorithm for explaining the second embodiment.
In this case, steps corresponding to those in FIG. 7 in the first embodiment described above are denoted by the same reference numerals, and redundant description is omitted. Also in this embodiment, the electrode position, the eyeball radius are estimated, and the current density initial value ^IBUF is set by the same calibration process A described above.
The difference between the second embodiment and the first embodiment is that the processing is closed in the process in each eyeball position measurement step E (E1, E2, E3,...). Rather, the current position I ^BUF converged in the previous step E is used to estimate the eye position of the stroke. The validity of such an algorithm depends on the fact that the change in current density is slow in the first place, and the magnitude of the current density does not change significantly between two consecutive sections. That is, the current density estimated using the potential in the previous stage was estimated from the previous stage to estimate the eye position of the stage, on the assumption that the current density in the stage was almost the same as the current density in the stage. The current density is used.

The A process of calibration is the same as that shown in FIG. 9, but “0” is first introduced at time t, and the potential difference signal observed at the electrode is acquired in step S5. In step S6, the previous section is acquired. Based on the idea of least square estimation, θ and ξ that minimize the evaluation function E (θ, ξ; I ^BUF ) are obtained by combining the current estimated value I ^BUF obtained from ^Are stored as i-th values θ ^BUF _i and ξ ^BUF _i (step S7). This value is the estimated eyeball position at that time. After that, after waiting the Δt based on the output of the timer (step S8), and increments the counter value (step S9), and it is determined whether or not exceeded the predetermined number N _i (step S10). When the counter value i is less than or equal to _{N i,} the control returns to step S5, step S10 is repeated from step S5. When the counter value i exceeds _{N i} combines _{N i} number of estimates theta ^BUF _i obtained so far, and xi] ^BUF _i, a potential difference value at that time, the evaluation function E (I; {θ The least square estimate of the current density I that minimizes ^BUF _i }, {ξ ^BUF _i }) is calculated (step S11), and the current density estimate I ^BUF is obtained (step 12). When the estimated value of the current density is obtained in this way, the counter i is reset (set to “0”), and the process returns to step S5 again. As can be seen from this, the estimated value I ^BUF of the current density obtained in this iterative calculation is used for eyeball position estimation in the next cycle. This point is different from the first embodiment in which both the current density and the eyeball position are estimated based on the data in the same section. As described above, the validity of the second embodiment is based on the fact that the magnitude of the current density hardly changes during successive intervals.

According to the eyeball measurement method according to the present invention described above, the measurement subject can only perform the gaze of the reference point, and thereafter can perform the measurement only with the electrode attached. As described at the beginning, the eyeball By avoiding the video camera and the incidental of the optical system, etc., the physical and mental burdens of the measurement subject, such as obstructing motion and visual field, are significantly reduced. Rather, it can be applied to various measurement purposes.
Despite the use of the EOG method in which measurement by an eyeball observation device such as a video camera or a light irradiation device is not performed in this way, in this case, for example, drift in the potential difference from the sensor data, that is, the potential difference from the electrode due to fluctuations in the current density of the eyeball etc. Instability due to is avoided.
That is, in the present invention, using the fact that the time scale of drift variation is sufficiently long (from 100 times) larger than the change in eyeball position, using the EM algorithm, Since the converged estimated amount is acquired and the eyeball position (eyeball position) is measured, measurement with high accuracy can be performed.

In addition, since a potential difference measured with a plurality of electrodes attached to the vicinity of the eyeball can be transmitted as an electric signal after being amplified and converted into a digital signal, it is possible to provide a measurement device that is not easily affected by noise. it can. And as a preferable form, since a wireless transmission technique can be utilized, the eye movement measuring device which is not restrict | limited at all by the motion of a measurement subject can be provided.
In particular, when radio is used for signal transmission, the movement of the measurement subject is completely free, and the eye movement of the subject can be measured for 24 hours as long as power can be supplied. Therefore, the electrical output obtained by the movement of the eyeball position can be used in many ways, such as control of a care machine, and its social and industrial contributions are great.

  Needless to say, the present invention is not limited to the above-described embodiments, and various forms are included without departing from the gist of the present invention described in the claims.

It is a figure which shows the battery model of an eyeball. It is a figure which shows the setting conditions of an eyeball position. It is explanatory drawing of the arrangement conditions of an electrode. It is a figure which shows the relationship between an electrode arrangement | positioning condition and the error of a gaze direction estimation. It is a figure which shows the relationship between an electrode arrangement | positioning pattern and the error of simultaneous estimation of a current density and an electrode position. It is explanatory drawing of a gaze direction and current density simultaneous estimation. It is a block block diagram which shows the example of embodiment of this invention. It is a block block diagram which shows the other embodiment (modification) of this invention. It is a flowchart of the embodiment of this invention. It is a flowchart for demonstrating the initial value work (calibration) in the example of embodiment of this invention. It is a time chart for demonstrating the whole operation | movement of the embodiment of this invention. It is a time chart for demonstrating the operation | work of the other embodiment of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Eye, 2, 22 ... Electrode, 3 ... Signal conversion circuit, 4 ... Data processing circuit, 5 ... Central processing unit (CPU), 6 ... Timer, ... RAM, 8 ... ROM, 9 ... monitor, 11 ... transmission circuit, 12 ... reception circuit, 21 ... eyeball, 23 ... eye

Claims (4)

The potential detected by a plurality of electrodes of at least three or more points mounted around the eye is appropriately combined to extract the difference voltage,
Using the information on the plurality of extracted potential differences, the current density flowing from the retina side to the cornea side of the eyeball or the magnitude of the charge considered to be charged on the retina and cornea in the model in which the eyeball is regarded as a battery, and the eyeball Estimate the position, the position where the electrode was applied, and the size of the eyeball,
Based on the potential difference extracted after that using the estimated electrode application position and eyeball radius,
The eyeball position and the current density or the retina and cornea are charged using an EM (Expectation and Maximization) algorithm method that alternately repeats the estimation of the eyeball position and the estimation of the current density over a plurality of time cycles. A method for measuring an eyeball position, comprising estimating an amount of electric charge that is considered to be.   A potential difference obtained from the measured potential is transmitted as a digital signal. In a processor and a data processing circuit connected to the receiving circuit, the calculation is performed for estimating the eyeball position and estimating the current density. The eyeball position measuring method according to claim 1, wherein the eyeball position measuring method is performed. A plurality of at least three or more electrodes mounted around the eye;
A signal conversion circuit for detecting a potential difference between a plurality of electrodes from any combination between the electrodes;
From the potential difference extracted from the signal conversion circuit, the current density flowing from the retina side to the cornea side of the eyeball in the model where the electrode attachment position and the eyeball are regarded as a battery, or the magnitude of the charge considered to be charged in the retina and cornea A processor and a data processing circuit for estimating the eyeball position by repetitive calculation;
In the calculation of the data processing circuit, an eye position measurement apparatus characterized in that the eye position estimation and the current density estimation are calculated and calculated over a plurality of cycles by an EM (Expectation and Maximization) algorithm method.   The transmission circuit according to claim 3, further comprising: a transmission circuit that converts a potential difference obtained from the measured potential into a digital signal and transmits the digital signal; and a reception circuit that receives a signal transmitted from the transmission circuit by wire or wirelessly. The eyeball position measurement apparatus described.

Images