JPH0939603A - Dozing judgement device - Google Patents

Dozing judgement device

Info

Publication number
JPH0939603A
JPH0939603A JP7195491A JP19549195A JPH0939603A JP H0939603 A JPH0939603 A JP H0939603A JP 7195491 A JP7195491 A JP 7195491A JP 19549195 A JP19549195 A JP 19549195A JP H0939603 A JPH0939603 A JP H0939603A
Authority
JP
Japan
Prior art keywords
eye
time
driver
objective
dozing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
JP7195491A
Other languages
Japanese (ja)
Inventor
Makoto Nishida
誠 西田
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Toyota Motor Corp
Original Assignee
Toyota Motor Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Toyota Motor Corp filed Critical Toyota Motor Corp
Priority to JP7195491A priority Critical patent/JPH0939603A/en
Publication of JPH0939603A publication Critical patent/JPH0939603A/en
Pending legal-status Critical Current

Links

Classifications

    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02TCLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00Road transport of goods or passengers
    • Y02T10/80Technologies aiming to reduce greenhouse gasses emissions common to all road transportation technologies
    • Y02T10/84Data processing systems or methods, management, administration

Landscapes

  • Emergency Alarm Devices (AREA)
  • Auxiliary Drives, Propulsion Controls, And Safety Devices (AREA)
  • Image Processing (AREA)

Abstract

PROBLEM TO BE SOLVED: To provide a dozing alarm suitable for feeling of a driver by providing an eye closing time predicting means to predict continuous time of an eye closing state based on a plural number of data of past eye closing time and a judgement means to judge it as dozing at the time when continuous time of a predicted following eye closing state exceeds specified time. SOLUTION: A picture image processing device 12 detects objective templates (objective face template, objective eye neighbourhood template, objective eye template) of an objective driver by correlative computation by using a standard template stored in a memory 14. At the time of forming the objective templates, a face allocation data stored in the memory 14. The formed objective templates are stored in the memory 14, and change of opening and closing states of eyes of the driver is monitored by correlative computation from a face picture image of the driver by using these objective templates. This result is supplied to an ECU 16, existence of an abnormal state of the driver is judged, and an alarm is raised at the abnormal time.

Description

【発明の詳細な説明】Detailed Description of the Invention

【0001】[0001]

【発明の属する技術分野】本発明は、居眠り判定装置に
関し、被験者の眼の開閉状態から被験者の居眠り判定を
行う居眠り判定装置に関する。
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a drowsiness determination device, and more particularly to a drowsiness determination device for determining a subject's drowsiness based on whether the subject's eyes are open or closed.

【0002】[0002]

【従来の技術】従来より運転者の顔を撮像した顔画像か
ら運転者の瞬きを検出し、この瞬きの状態から運転者の
居眠り等を検出する装置が提案されている。例えば、特
開平6−32154号公報には、運転者の目の縦幅を検
出し、この目の縦幅の最大値及び最小値から閾値を設定
して瞬きを検出し、瞬きの時間変化から閉眼状態が所定
時間以上継続したとき運転者が居眠り状態にあると判定
している。
2. Description of the Related Art Conventionally, there has been proposed a device for detecting the blink of the driver from a face image obtained by capturing the face of the driver and detecting the drowsiness of the driver or the like from the state of the blink. For example, in Japanese Unexamined Patent Publication No. 6-32154, the vertical width of a driver's eyes is detected, a threshold is set from the maximum value and the minimum value of the vertical width of the driver's eyes to detect blinks, and the temporal change of the blinks is detected. It is determined that the driver is in a dozing state when the eyes closed state continues for a predetermined time or longer.

【0003】[0003]

【発明が解決しようとする課題】例えば30秒等の所定
時間当りの平均瞬き閉眼時間Teav と、意識レベルLc
とは図8に示す関係にある。このため、従来は平均瞬き
閉眼時間Teav が閾値TH未満のとき居眠りと判定して
いた。
Mean blink closing time Teav per predetermined time such as 30 seconds and consciousness level Lc
And have the relationship shown in FIG. For this reason, conventionally, when the average blink-eye closing time Teav is less than the threshold value TH, it is determined to be dozing.

【0004】ここで、車両運転中の覚醒状態から居眠り
状態へと変化する場合は、瞬き閉眼時間Te 、意識レベ
ルLcは図9に示すように変化する。図8に示す意識レ
ベルLcが眠いからかなり眠いまでの間は平均瞬き閉眼
時間Teav の変化が大きいのに対し、かなり眠いから居
眠りまでの間は平均瞬き閉眼時間Teav の変化が小さ
い。
Here, when the awake state during driving of the vehicle is changed to the dozing state, the blinking eye closing time Te and the consciousness level Lc are changed as shown in FIG. While the consciousness level Lc shown in FIG. 8 changes significantly from the sleepy state to the sleepy state, the average blinking eye closing time Teav changes greatly, while the average blinking eye closing time Teav changes from the considerably sleepy state to the dozing state small.

【0005】これはかなり眠いから居眠りまでの過程で
は、閉眼時間が比較的長い1〜2秒の瞬きが多発し、平
均値であるTeav も微増するが、居眠りは突発的に閉眼
時間が2秒以上の瞬きが発生したときにおちいると考え
られるので、平均値であるTeav は微増にとどまるため
である。従って、従来、閾値THは1.8秒程度に設定
している。
[0005] In the process from quite drowsy to dozing, blinks frequently occur for 1 to 2 seconds with a relatively long eye-closing time, and Teav, which is the average value, slightly increases, but during dozing, the eye-closing time is suddenly 2 seconds. This is because the average value Teav is only slightly increased because it is considered that the blink occurs when the above blink occurs. Therefore, conventionally, the threshold value TH is set to about 1.8 seconds.

【0006】従来は、平均瞬き閉眼時間Teav が例えば
1.8秒の閾値THを超えた時点で居眠りと判定し、警
報を行っているが、この時点では運転者はかなり眠いだ
けで未だ居眠りをしているわけではないため、この警報
が運転者の感覚に合わず、わずらわしいという問題があ
った。
Conventionally, when the average blink closing time Teav exceeds a threshold TH of 1.8 seconds, for example, it is determined that the driver is dozing and an alarm is issued, but at this time, the driver is quite drowsy and still drowsy. However, there is a problem that this warning does not match the driver's feeling and is troublesome.

【0007】ところで、閾値THを2秒と設定すること
も考えられるが、この場合には運転者が居眠りを始め、
閉眼から2秒経過して警報が発せられるため、警報が遅
いという問題がある。本発明は、上記の点に鑑みなされ
たもので、次の閉眼状態の継続時間を予測し、この予測
された次の閉眼時間により居眠り判定を行うことによ
り、居眠りを精度良く判定し、運転者の感覚に合った居
眠りの警報を行うことのできる居眠り判定装置を提供す
ることを目的とする。
By the way, although it is possible to set the threshold value TH to 2 seconds, in this case, the driver starts to fall asleep,
There is a problem that the alarm is delayed because the alarm is issued 2 seconds after the eyes are closed. The present invention has been made in view of the above point, predicts the duration of the next eye-closed state, by performing a doze determination by this predicted next eye-closing time, to accurately determine the doze, the driver It is an object of the present invention to provide a drowsiness determination device capable of issuing a drowsiness alarm that matches the feeling of.

【0008】[0008]

【課題を解決するための手段】請求項1に記載の発明
は、図1(A)に示す如く、検出手段M1で被験者の眼
の開閉状態を検出し、閉眼状態が所定時間を超えて継続
したとき居眠り判定を行う居眠り判定装置において、過
去の閉眼時間の複数のデータに基づいて、次の閉眼状態
の継続時間を予測する閉眼時間予測手段M2と、上記予
測された次の閉眼状態の継続時間が所定時間を超えたと
き居眠りと判定する判定手段M3とを有する。
According to the invention described in claim 1, as shown in FIG. 1 (A), the open / closed state of the eye of the subject is detected by the detecting means M1, and the closed eye state continues for a predetermined time or longer. In the drowsiness determination device that performs the drowsiness determination, the eye-closing time predicting unit M2 that predicts the duration of the next eye-closed state based on a plurality of data of the past eye-closing time, and the continuation of the predicted next eye-closed state. And a determining unit M3 that determines to fall asleep when the time exceeds a predetermined time.

【0009】このため、居眠りに入った場合は、予測さ
れる次の閉眼時間は例えば2秒以上の大きな値となり、
かなり眠い状態では予測される次の閉眼時間は2秒未満
の値となるため、閾値とする所定時間を例えば2秒とす
ることにより、居眠りを精度良く判定でき、かつ、居眠
りが始まる前に判定できる。
Therefore, when the person falls asleep, the predicted next eye-closing time becomes a large value of, for example, 2 seconds or more,
In the case of being quite sleepy, the predicted next eye-closing time will be a value of less than 2 seconds, so by setting the predetermined time as the threshold value to, for example, 2 seconds, it is possible to accurately determine drowsiness and to determine before drowsiness begins. it can.

【0010】請求項2に記載の発明は、図1(B)に示
す如く、検出手段M1で被験者の眼の開閉状態を検出
し、閉眼状態が所定時間を超えて継続したとき居眠り判
定を行う居眠り判定装置において、過去の閉眼時間の複
数のデータに基づいて、次の閉眼時期及び閉眼状態の継
続時間を予測する閉眼時期及び時間予測手段M4と、上
記予測された次の閉眼状態の継続時間が上記所定時間を
超えたとき居眠りと判定する判定手段M3と、居眠りと
判定されたとき次の閉眼時期に警報を行う警報制御手段
M5とを有する。
According to the second aspect of the present invention, as shown in FIG. 1B, the detection means M1 detects the open / closed state of the subject's eyes, and when the closed eye state continues for more than a predetermined time, a drowsiness determination is performed. In the doze determination device, an eye-closing time and time predicting means M4 for predicting a next eye-closing time and a duration of the eye-closed state based on a plurality of pieces of data of the past eye-closing time, and the predicted duration of the next eye-closed state. Has a determination means M3 for determining that the subject has fallen asleep and the alarm control means M5 for issuing an alarm at the next eye closing time when it has been determined that the subject has fallen asleep.

【0011】このため、請求項1と同様に居眠りを精度
良く判定でき、かつ居眠りが始まる前に判定できると共
に、居眠りが始まるタイミングで警報を行うことができ
るため、運転者の感覚に合った居眠り警報を行うことが
できる。
Therefore, as in the first aspect of the present invention, it is possible to accurately determine the doze, and it is possible to make a determination before the dozing starts, and it is possible to give an alarm at the timing when the dozing starts, so that the dozing suitable for the driver's feeling can be obtained. An alarm can be given.

【0012】[0012]

【発明の実施の形態】図2は本発明のブロック図を示
す。同図中、運転者を撮像するカメラ10が車両の所定
位置に設けられ、運転者の顔画像を撮像する。カメラ1
0には画像処理装置12が接続されており、得られた運
転者の顔画像を画像処理装置12に供給する。画像処理
装置12にはA/D変換器、正規化回路及び相関演算回
路を備えており、入力された画像信号をデジタル信号に
変換し、更に濃淡正規化処理を行う。画像処理装置12
はメモリ14が接続されている。メモリ14には標準テ
ンプレート(基準顔テンプレート、標準目近傍テンプレ
ート、標準目テンプレート)及び眉毛や目などの顔要素
の配置データが予め格納されている。配置データの内容
は、「眉毛、目は横方向(x方向)に平行して存在す
る」、「目は眉毛の下方にある」等である。
2 shows a block diagram of the present invention. In the figure, a camera 10 for picking up an image of the driver is provided at a predetermined position of the vehicle to take a face image of the driver. Camera 1
An image processing device 12 is connected to 0, and the obtained face image of the driver is supplied to the image processing device 12. The image processing device 12 includes an A / D converter, a normalization circuit, and a correlation calculation circuit, converts the input image signal into a digital signal, and further performs a grayscale normalization process. Image processing device 12
Is connected to the memory 14. The memory 14 stores in advance standard templates (standard face template, standard eye vicinity template, standard eye template) and arrangement data of face elements such as eyebrows and eyes. The contents of the arrangement data are “eyebrows, eyes exist in parallel to the lateral direction (x direction)”, “eyes are below eyebrows”, and the like.

【0013】画像処理装置12ではメモリ14に格納さ
れた標準テンプレートを用いて対象運転者の対象テンプ
レート(対象顔テンプレート、対象目近傍テンプレー
ト、対象目テンプレート)を相関演算により検出する。
対象テンプレートを作成するに際し、メモリ14に格納
された顔配置データが用いられる。作成された対象テン
プレート22はメモリ14に格納され、この対象テンプ
レートを用いて運転者の顔画像から運転者の目の開閉状
態の変化を相関演算により監視する。この結果はECU
16に供給され、運転者の異常状態の有無が判定され、
異常時には警報が出される。
The image processing apparatus 12 detects the target template (target face template, target eye vicinity template, target eye template) of the target driver by correlation calculation using the standard template stored in the memory 14.
The face arrangement data stored in the memory 14 is used when creating the target template. The created target template 22 is stored in the memory 14, and changes in the open / closed state of the eyes of the driver are monitored by correlation calculation from the face image of the driver using this target template. This result is ECU
16 to determine whether the driver is in an abnormal state,
An alarm is issued in the event of an abnormality.

【0014】図3はECU16が実行する居眠り判定処
理のフローチャートを示す。この処理は所定時間毎に繰
り返し実行される。同図中、ステップS10では予測用
マップをロードし、ステップS12で瞬きカウンタiを
0にリセットする。次にステップS14で瞬きデータを
取得する。ここで、瞬きデータとは、図4に示す如く、
今回の瞬きの閉眼時間Te(i)と、前回の瞬きとの間
隔である瞬き間隔Ie(i)とである。次にステップS
16で瞬きカウンタiを1だけインクリメントし、ステ
ップS18で瞬きカウンタiが所定値N−1未満か否か
を判別する。i<N−1の場合はステップS14に進ん
で瞬きデータの取得を繰り返す。
FIG. 3 shows a flow chart of the drowsiness determination process executed by the ECU 16. This process is repeatedly executed at predetermined time intervals. In the figure, the prediction map is loaded in step S10, and the blink counter i is reset to 0 in step S12. Next, in step S14, blink data is acquired. Here, the blink data means, as shown in FIG.
It is the eye-closure time Te (i) of this blink and the blink interval Ie (i) which is the interval between the last blink. Next, step S
In step 16, the blink counter i is incremented by 1, and in step S18 it is determined whether or not the blink counter i is less than a predetermined value N-1. If i <N−1, the process proceeds to step S14 and the acquisition of blink data is repeated.

【0015】i≧N−1となるとステップS20で瞬き
データを取得した後、ステップS22に進んで瞬きデー
タセット{Te(i),Te(i−1),…,Te(i
−N+1)}及び{Ie(i),Ie(i−1),…,
Ie(i−N+1)}を作成する。この後、次のステッ
プS24で状態ベクトルVTe(i),VIe(i)か
ら次の瞬きデータTe(i+1),Ie(i+1)を予
測する。この後、ステップS26で予測された次の瞬き
の閉眼時間Te(i+1)が所定の閾値THEを超えて
いるかどうかを判別し、Te(i+1)≦THEであれ
ばステップS27でiを1だけインクリメントしてステ
ップS20に進み、瞬きデータの取得及び次の瞬きデー
タの予測を繰り返す。ここで、閾値THEは例えば2秒
に相当する値とされている。Te(i+1)>THEで
あればステップS28に進み、予測された次の瞬きまで
の間隔Ie(i)だけ時間経過後、警報装置18を用い
て居眠り警報を行って処理を終了する。つまり、この処
理は運転中は全時間動作するものであり、イグニッショ
ンオフ、又は居眠り警報を行ったときに終了する。上記
のステップS24が閉眼時間予測手段M2又は閉眼時期
及び閉眼時間予測手段M4に対応し、ステップS26が
判定手段M3に対応し、ステップS28が警報制御手段
M5に対応する。
When i ≧ N−1, the blink data is acquired in step S20, and then the process proceeds to step S22, in which blink data set {Te (i), Te (i−1), ..., Te (i
-N + 1)} and {Ie (i), Ie (i-1), ...,
Ie (i-N + 1)} is created. Then, in the next step S24, the next blink data Te (i + 1), Ie (i + 1) is predicted from the state vectors VTe (i), VIe (i). Thereafter, it is determined whether or not the eye blink closing time Te (i + 1) predicted for the next blink in step S26 exceeds a predetermined threshold value THE, and if Te (i + 1) ≦ THE, i is incremented by 1 in step S27. Then, the process proceeds to step S20, and the acquisition of blink data and the prediction of the next blink data are repeated. Here, the threshold value THE is set to a value corresponding to, for example, 2 seconds. If Te (i + 1)> THE, the process proceeds to step S28, and after the elapse of the predicted interval Ie (i) until the next blink, the alarm device 18 is used to give a drowsiness alarm and the process ends. In other words, this process operates all the time during driving, and ends when the ignition is turned off or the doze alarm is given. The above step S24 corresponds to the eye closing time predicting means M2 or the eye closing timing and eye closing time predicting means M4, step S26 corresponds to the judging means M3, and step S28 corresponds to the alarm control means M5.

【0016】次に、ステップS24で実行する瞬きデー
タ予測ルーチンについて説明する。まず、カオス理論に
よる予測を説明する。i番目の瞬きの閉眼時間Te
(i),間隔Ie(i)をタケンスの理論に基づいて埋
め込みを行い、状態ベクトルVTe(i),VIe
(i) VTe(i)={Te(i),Te(i−1),…Te(i−N+1)} VIe(i)={Ie(i),Ie(i−1),…Ie(i−N+1)} を図5(A),(B)に示す状態空間上に変換する。こ
こでNはフラクタル次元解析により決定される埋め込み
次元であり、例えばN=2である。
Next, the blink data prediction routine executed in step S24 will be described. First, the prediction by the chaos theory will be explained. Eye closing time Te of the i-th blink
(I), the interval Ie (i) is embedded based on the Takens theory, and the state vectors VTe (i), VIe
(I) VTe (i) = {Te (i), Te (i-1), ... Te (i-N + 1)} VIe (i) = {Ie (i), Ie (i-1), ... Ie ( i-N + 1)} is transformed into the state space shown in FIGS. Here, N is an embedding dimension determined by fractal dimension analysis, and is N = 2, for example.

【0017】状態ベクトルVTe(i)がカオスである
ならば、VTe(i)の近傍にある状態VTe(j)と
VTe(i)とは、VTe(i)近傍の状態変化関数F
aのヤコビ行列DFaにより次の関係がある。なお、図
5(A)のFa(=〔Fa(i)、…、Fa(i−n+
1〕)は既に学習済みとする。 VTe(j+1)−VTe(i+1)=DFa(VTe(j)−VTe(i)) …(1)
If the state vector VTe (i) is chaotic, the states VTe (j) and VTe (i) in the vicinity of VTe (i) are the state change function F in the vicinity of VTe (i).
The Jacobian matrix DFa of a has the following relationship. Note that Fa (= [Fa (i), ..., Fa (i−n +) in FIG.
1]) is already learned. VTe (j + 1) -VTe (i + 1) = DFa (VTe (j) -VTe (i)) (1)

【0018】[0018]

【数1】 [Equation 1]

【0019】従って、VTe(i)の次の状態であるV
Te(i+1)は、 VTe(i+1)=DFa(VTe(j)−VTe(i))+VTe(j+1) …(2) と表わされ、この状態ベクトルVTe(i+1)の要素
である次の瞬きの閉眼時間Te(i+1)を予測するこ
とができる。
Therefore, the next state of VTe (i) is V
Te (i + 1) is expressed as VTe (i + 1) = DFa (VTe (j) −VTe (i)) + VTe (j + 1) (2), and the next blink which is an element of this state vector VTe (i + 1). The eye-closing time Te (i + 1) can be predicted.

【0020】同様に状態ベクトルVIe(i)がカオス
であるならば、VIe(i)の近傍にある状態VIe
(j)とVIe(i)とは、VIe(i)近傍の状態変
化関数Fbのヤコビ行列DFbにより次の関係がある。
なお、図5(B)のFbは既に学習済みとする。 VIe(j+1)−VIe(i+1)=DFb(VIe(j)−VIe(i)) …(3)
Similarly, if the state vector VIe (i) is chaotic, the state VIe in the vicinity of VIe (i) is
(J) and VIe (i) have the following relationship due to the Jacobian matrix DFb of the state change function Fb near VIe (i).
Note that Fb in FIG. 5B is already learned. VIe (j + 1) -VIe (i + 1) = DFb (VIe (j) -VIe (i)) (3)

【0021】[0021]

【数2】 [Equation 2]

【0022】従って、VIe(i)の次の状態であるV
Ie(i+1)は、 VIe(i+1)= DFbi(VIe(j)−VIe(i))+VIe(j+1) …(4) と表わされ、この状態ベクトルVIe(i+1)の要素
である次の瞬きの間隔Ie(i+1)を予測することが
できる。
Therefore, the next state of VIe (i) is V
Ie (i + 1) is represented as VIe (i + 1) = DFbi (VIe (j) -VIe (i)) + VIe (j + 1) (4), and the next blink which is an element of this state vector VIe (i + 1). The interval Ie (i + 1) can be predicted.

【0023】図6はカオス理論を用いた瞬きデータ予測
ルーチンのフローチャートを示す。同図中、ステップS
32ではステップS22で作成した瞬きデータセット
{Te(i),Te(i−1),…,Te(i−N+
1)},{Ie(i),Ie(i−1),…,Ie(i
−N+1)}を夫々状態ベクトルVTe(i),VIe
(i)とする。次にステップS34で図4(A),
(B)に示す状態変化関数Fa,Fb上のTe(i),
Ie(i)の近傍点Te(j),Ie(j)を設定す
る。ステップS36ではTe(i),Ie(i)夫々の
ヤコビ行列DFa,DFbを計算する。次にステップS
38で(2)式,(4)式を用いて状態ベクトルVTe
(i+1),VIe(i+1)夫々を計算し、この状態
ベクトルから次の瞬きの閉眼時間Te(i+1)及び瞬
き間隔Ie(i+1)夫々を得て処理を終了する。
FIG. 6 shows a flow chart of a blink data prediction routine using chaos theory. In FIG.
In 32, the blink data set {Te (i), Te (i-1), ..., Te (i-N +) created in step S22.
1)}, {Ie (i), Ie (i-1), ..., Ie (i
-N + 1)} are respectively state vectors VTe (i), VIe
(I). Next, in step S34, as shown in FIG.
Te (i) on the state change functions Fa and Fb shown in (B),
The neighboring points Te (j) and Ie (j) of Ie (i) are set. In step S36, the Jacobian matrices DFa and DFb of Te (i) and Ie (i) are calculated. Next, step S
38 using the equations (2) and (4), the state vector VTe
(I + 1) and VIe (i + 1) are calculated, and the eye closing time Te (i + 1) and blink interval Ie (i + 1) of the next blink are obtained from this state vector, and the process is terminated.

【0024】次に、線形予測法による予測を説明する。
この方法では次に示す線形予測式を用いる。 Te(i+1)=a0 Te(i)+a1 Te(i−1)… +aN-1 Te(i−N+1) =f(VTe(i)) …(5) Ie(i+1)=b0 Ie(i)+b1 Ie(i−1)… +bN-1 Ie(i−N+1) =g(VIe(i)) …(6) 上記(5),(6)式の係数a0 〜aN-1 ,b0 〜b
N-1 夫々は予め採取した閉眼時間データセット{V
0 ,…VxM-1 },Vxi=(xi ,…,xi-m1)、瞬
き間隔データセット{Vy0 ,…VyN-1 },Vyi=
(yi ,…,yi-m2)但し、xi,yi夫々はi番目の
瞬きの閉眼時間,瞬き間隔であり、N,Mは例えば10
00、m1,m2は例えば100である。各採取データ
を(5),(6)式に当てはめることにより次式の如く
なる。
Next, the prediction by the linear prediction method will be described.
This method uses the following linear prediction formula. Te (i + 1) = a 0 Te (i) + a 1 Te (i-1) ... + a N-1 Te (i−N + 1) = f (VTe (i)) (5) Ie (i + 1) = b 0 Ie (I) + b 1 Ie (i−1) ... + b N−1 Ie (i−N + 1) = g (VIe (i)) (6) The coefficients a 0 to a N of the above equations (5) and (6). -1 , b 0 ~ b
N-1 is the eye-closing time data set {V
x 0 , ... Vx M-1 }, Vxi = (x i , ..., x i-m1 ), blink interval data set {Vy 0 , ... Vy N-1 }, Vyi =
(y i , ..., y i-m2 ) where xi and yi are eye closing time and blink interval of the i-th blink, and N and M are, for example, 10
00, m1, and m2 are 100, for example. By applying each sampled data to the equations (5) and (6), the following equation is obtained.

【0025】 xi=a0 i-1 +a1 i-2 +…+am1-1i-m1(i
=0〜M−1) yi=b0 i-1 +b1 i-2 +…+bm2-1i-m2(i
=0〜N−1) ここで、Vxiのj成分Xi-j をXij、Vyiのj成分
i-j をyijと表わすと以下に示す如くなる。
Xi = a 0 x i-1 + a 1 x i-2 + ... + a m1-1 x i-m1 (i
= 0 to M-1) yi = b 0 y i-1 + b 1 y i-2 + ... + b m2-1 y i-m2 (i
= 0 to N−1) Here, the j component X ij of Vxi is represented as X ij , and the j component y ij of Vyi is represented as y ij , as follows.

【0026】 xi0=a0 i1+a1 i2+…+am1-1im1i0=b0 i1+b1 i2+…+bm2-1im2 ここで、ai i=0,m1-1)、bj j=0,m2-1)は次
の(7),(8)式で与えられる。
X i0 = a 0 x i1 + a 1 x i2 + ... + a m1-1 x im1 y i0 = b 0 y i1 + b 1 y i2 + ... + b m2-1 y im2 where a i ( i = 0 , ... m1-1 ), b j ( j = 0, ... m2-1 ) are given by the following equations (7) and (8).

【0027】[0027]

【数3】 (Equation 3)

【0028】図7は線形予測法を用いたデータ予測ルー
チンのフローチャートを示す。同図中、ステップS42
ではステップS22で作成した瞬きデータセット{Te
(i),Te(i−1),…,Te(i−N+1)},
{Ie(i),Ie(i−1),…,Ie(i−N+
1)}を夫々状態ベクトルVTe(i),VIe(i)
とする。次にステップS44で(5),(6)式を用い
て次の瞬きの閉眼時間Te(i+1)及び瞬き間隔Ie
(i+1)を計算して処理を終了する。
FIG. 7 shows a flow chart of a data prediction routine using the linear prediction method. In the figure, step S42
Then, the blink data set created in step S22 {Te
(I), Te (i-1), ..., Te (i-N + 1)},
{Ie (i), Ie (i-1), ..., Ie (i-N +
1)} are state vectors VTe (i) and VIe (i), respectively.
And Next, in step S44, using the expressions (5) and (6), the eye blink closing time Te (i + 1) and the blink interval Ie of the next blink are calculated.
(I + 1) is calculated, and the process ends.

【0029】ここで、居眠りに入った場合は予測される
閉眼時間Te(i+1)は例えば2秒以上の大きな値と
なり、かなり眠い状態では予測される次の閉眼時間Te
(i+1)は2秒未満の値となる。このため、閾値TH
Eを例えば2秒とすることにより、かなり眠い状態を居
眠りと判定することなく、居眠りを精度良く判定でき
る。また、この判定は居眠りが始まる前に行われる。更
に、居眠りが始まるタイミングで警報を行うため、運転
者の感覚に合った居眠り警報を行うことができる。
When the person falls asleep, the predicted eye-closing time Te (i + 1) becomes a large value of, for example, 2 seconds or more, and in a considerably sleepy state, the next eye-closing time Te (i + 1) predicted.
(I + 1) has a value of less than 2 seconds. Therefore, the threshold TH
By setting E to 2 seconds, for example, it is possible to accurately determine the doze without determining that the state is quite sleepy. Also, this determination is made before the start of dozing. Further, since the alarm is issued at the timing when the drowsiness starts, it is possible to give the drowsiness alarm that suits the driver's sense.

【0030】[0030]

【発明の効果】上述の如く、請求項1に記載の発明は、
被験者の眼の開閉状態を検出し、閉眼状態が所定時間を
超えて継続したとき居眠り判定を行う居眠り判定装置に
おいて、過去の閉眼時間の複数のデータに基づいて、次
の閉眼状態の継続時間を予測する閉眼時間予測手段と、
上記予測された次の閉眼状態の継続時間が所定時間を超
えたとき居眠りと判定する判定手段とを有する。
As described above, the invention according to claim 1 is
Detecting the open / closed state of the subject's eyes, in a dozing determination device that performs a dozing determination when the closed state continues for more than a predetermined time, based on multiple data of the past closed eye time, the duration of the next closed eye state. Eye-closing time predicting means for predicting,
And a determination unit that determines that the subject is asleep when the predicted duration of the next eye-closed state exceeds a predetermined time.

【0031】このため、居眠りに入った場合は、予測さ
れる次の閉眼時間は例えば2秒以上の大きな値となり、
かなり眠い状態では予測される次の閉眼時間は2秒未満
の値となるため、閾値とする所定時間を例えば2秒とす
ることにより、居眠りを精度良く判定でき、かつ、居眠
りが始まる前に判定できる。
Therefore, when the person falls asleep, the predicted next eye-closing time becomes a large value of, for example, 2 seconds or more,
In the case of being quite sleepy, the predicted next eye-closing time will be a value of less than 2 seconds, so by setting the predetermined time as the threshold value to, for example, 2 seconds, it is possible to accurately determine drowsiness and to determine before drowsiness begins. it can.

【0032】請求項2に記載の発明は、被験者の眼の開
閉状態を検出し、閉眼状態が所定時間を超えて継続した
とき居眠り判定を行う居眠り判定装置において、過去の
閉眼時間の複数のデータに基づいて、次の閉眼時期及び
閉眼状態の継続時間を予測する閉眼時期及び時間予測手
段と、上記予測された次の閉眼状態の継続時間が上記所
定時間を超えたとき居眠りと判定する判定手段と、居眠
りと判定されたとき次の閉眼時期に警報を行う警報制御
手段とを有する。
According to a second aspect of the present invention, in a drowsiness determination device that detects an eye open / closed state of a subject and determines a drowsiness when the eye closed state continues for more than a predetermined time, a plurality of data of past eye closing times are stored. Based on, the eye closing timing and time predicting means for predicting the next eye closing time and the duration of the eye closing state, and the determining means for determining dozing when the predicted duration of the next eye closing state exceeds the predetermined time And alarm control means for issuing an alarm at the next eye closing time when it is determined that the subject is dozing.

【0033】このため、請求項1と同様に居眠りを精度
良く判定でき、かつ居眠りが始まる前に判定できると共
に、居眠りが始まるタイミングで警報を行うことができ
るため、運転者の感覚に合った居眠り警報を行うことが
できる。
Therefore, as in the first aspect, it is possible to accurately determine the dozing, and it is possible to make a determination before the dozing starts, and it is possible to give an alarm at the timing when the dozing starts, so that the dozing that suits the driver's sensation can be performed. An alarm can be given.

【図面の簡単な説明】[Brief description of drawings]

【図1】本発明の原理図である。FIG. 1 is a principle diagram of the present invention.

【図2】本発明装置のブロック図である。FIG. 2 is a block diagram of the device of the present invention.

【図3】本発明の居眠り判定処理のフローチャートであ
る。
FIG. 3 is a flowchart of a dozing determination process according to the present invention.

【図4】閉眼時間と瞬き間隔を示す図である。FIG. 4 is a diagram showing eye-closing time and blink interval.

【図5】カオス理論による居眠り判定を説明するための
図である。
FIG. 5 is a diagram for explaining the doze determination based on the chaos theory.

【図6】瞬き予測ルーチンのフローチャートである。FIG. 6 is a flowchart of a blink prediction routine.

【図7】瞬き予測ルーチンのフローチャートである。FIG. 7 is a flowchart of a blink prediction routine.

【図8】平均閉眼時間と意識レベルとの関係を示す図で
ある。
FIG. 8 is a diagram showing a relationship between an average eye closing time and a consciousness level.

【図9】平均閉眼時間と意識レベルとの関係を示す図で
ある。
FIG. 9 is a diagram showing a relationship between an average eye-closing time and a consciousness level.

【符号の説明】[Explanation of symbols]

10 カメラ 12 画像処理装置 14 メモリ 15 ECU 18 警報装置 M1 検出手段 M2 閉眼時間予測手段 M3 判定手段 M4 閉眼時期及び閉眼時間予測手段 M5 警報制御手段 10 camera 12 image processing device 14 memory 15 ECU 18 alarm device M1 detecting means M2 eye closing time predicting means M3 determining means M4 eye closing timing and eye closing time predicting means M5 alarm control means

Claims (2)

【特許請求の範囲】[Claims] 【請求項1】 被験者の眼の開閉状態を検出し、閉眼状
態が所定時間を超えて継続したとき居眠り判定を行う居
眠り判定装置において、 過去の閉眼時間の複数のデータに基づいて、次の閉眼状
態の継続時間を予測する閉眼時間予測手段と、 上記予測された次の閉眼状態の継続時間が所定時間を超
えたとき居眠りと判定する判定手段とを有することを特
徴とする居眠り判定装置。
1. A drowsiness determination device for detecting a subject's eye open / closed state and performing a drowsiness determination when the eye closed state continues for more than a predetermined time, in which a next eye closing is performed based on a plurality of data of past eye closing times. An eye-sleeping determination device, comprising: an eye-closed time predicting unit that predicts a duration of a state; and a determining unit that determines that the user is asleep when the duration of the predicted next eye-closed state exceeds a predetermined time.
【請求項2】 被験者の眼の開閉状態を検出し、閉眼状
態が所定時間を超えて継続したとき居眠り判定を行う居
眠り判定装置において、 過去の閉眼時間の複数のデータに基づいて、次の閉眼時
期及び閉眼状態の継続時間を予測する閉眼時期及び時間
予測手段と、 上記予測された次の閉眼状態の継続時間が上記所定時間
を超えたとき居眠りと判定する判定手段と、 居眠りと判定されたとき次の閉眼時期に警報を行う警報
制御手段とを有することを特徴とする居眠り判定装置。
2. A doze determination device for detecting a subject's eye open / closed state and performing a drowsiness determination when the eye closed state continues for more than a predetermined time, in which a next eye closing is performed based on a plurality of data of past eye closing times. Eye closing time and time predicting means for predicting the duration and the duration of the eye-closed state, determination means for determining that the duration of the predicted next eye-closing state exceeds the predetermined time, and determination of dozing, and determination of dozing And a warning control means for warning at the next eye closing time.
JP7195491A 1995-07-31 1995-07-31 Dozing judgement device Pending JPH0939603A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP7195491A JPH0939603A (en) 1995-07-31 1995-07-31 Dozing judgement device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP7195491A JPH0939603A (en) 1995-07-31 1995-07-31 Dozing judgement device

Publications (1)

Publication Number Publication Date
JPH0939603A true JPH0939603A (en) 1997-02-10

Family

ID=16341978

Family Applications (1)

Application Number Title Priority Date Filing Date
JP7195491A Pending JPH0939603A (en) 1995-07-31 1995-07-31 Dozing judgement device

Country Status (1)

Country Link
JP (1) JPH0939603A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6571002B1 (en) * 1999-05-13 2003-05-27 Mitsubishi Denki Kabushiki Kaisha Eye open/close detection through correlation
JP2008035964A (en) * 2006-08-02 2008-02-21 Toyota Motor Corp Apparatus and program for determination of sleepiness
JP2009018091A (en) * 2007-07-13 2009-01-29 Toyota Motor Corp Doze detector
WO2009054039A1 (en) * 2007-10-22 2009-04-30 Pioneer Corporation Volume controller, volume control method, volume control program, and recording medium
JP2011086186A (en) * 2009-10-16 2011-04-28 Aisin Seiki Co Ltd Sleepiness determination apparatus
WO2013031138A1 (en) 2011-08-26 2013-03-07 Canon Kabushiki Kaisha Blink measurement device, method therefor and program
US8537000B2 (en) 2007-01-24 2013-09-17 Toyota Jidosha Kabushiki Kaisha Anti-drowsing device and anti-drowsing method

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6571002B1 (en) * 1999-05-13 2003-05-27 Mitsubishi Denki Kabushiki Kaisha Eye open/close detection through correlation
JP2008035964A (en) * 2006-08-02 2008-02-21 Toyota Motor Corp Apparatus and program for determination of sleepiness
US8537000B2 (en) 2007-01-24 2013-09-17 Toyota Jidosha Kabushiki Kaisha Anti-drowsing device and anti-drowsing method
JP2009018091A (en) * 2007-07-13 2009-01-29 Toyota Motor Corp Doze detector
WO2009054039A1 (en) * 2007-10-22 2009-04-30 Pioneer Corporation Volume controller, volume control method, volume control program, and recording medium
JP2011086186A (en) * 2009-10-16 2011-04-28 Aisin Seiki Co Ltd Sleepiness determination apparatus
WO2013031138A1 (en) 2011-08-26 2013-03-07 Canon Kabushiki Kaisha Blink measurement device, method therefor and program

Similar Documents

Publication Publication Date Title
CN107977638B (en) Video monitoring alarm method, device, computer equipment and storage medium
EP3929824A2 (en) Robust multimodal sensor fusion for autonomous driving vehicles
US20200334477A1 (en) State estimation apparatus, state estimation method, and state estimation program
Sigari et al. A driver face monitoring system for fatigue and distraction detection
Jodoin et al. Behavior subtraction
CN110297536B (en) Control method and electronic equipment
WO2008090451A2 (en) Anti-drowsing device and anti-drowsing method
WO2008127465A1 (en) Real-time driving danger level prediction
CN1278090A (en) Method for sensing driver being sleepy for driver sleepy alarming system
US20090177602A1 (en) Systems and methods for detecting unsafe conditions
CN111936990A (en) Method and device for waking up screen
JP2008099884A (en) Condition estimating apparatus
CN111488855A (en) Fatigue driving detection method, device, computer equipment and storage medium
KR20190083155A (en) Apparatus and method for detecting state of vehicle driver
JP2008065776A (en) Doze detection device and doze detection method
JPH0939603A (en) Dozing judgement device
JP7328089B2 (en) Eye closure determination device
CN113581209A (en) Driving assistance mode switching method, device, equipment and storage medium
CN115937830A (en) Special vehicle-oriented driver fatigue detection method
JP2002279410A (en) Detector for dozing
JP2011003076A (en) Risk recognition system
KR20190056873A (en) Apparatus for detecting object using multi neural network and method thereof, and method for learning detection of object
JP5050794B2 (en) Sleepiness detection device, sleepiness detection method
EP3680812A1 (en) Method for classifying an image taken by a sensor
US10945651B2 (en) Arousal level determination device

Legal Events

Date Code Title Description
LAPS Cancellation because of no payment of annual fees