EP1280118B1 - Method for determining road traffic states - Google Patents

Abstract

The method involves acquiring ambient noise with a microphone arranged on a road, analyzing the ambient noise and associating it with predefined traffic conditions. The ambient noise from the microphone is passed to an evaluation unit as a sound pattern from which several sample values are generated and used to produce signals representing the levels in various frequency bands.

Description

The present invention relates to a method of detection of road traffic conditions from ambient noise a street according to the preamble of claim 1.

Z1

Sporadischer Verkehr

Z2

Mässiger Verkehr

Z3

Dichter Verkehr

Z4

Stau / "Stop and Go"

The increase in road traffic creates unpleasant traffic obstructions, which have the effect of "stop and go" or even congestion. Various measurement variables can be used for the amount of road traffic, such as the river as the number of vehicles per unit of time at a point on a road in each direction or the density as the number of vehicles on a lane section. Traffic conditions can, for example, be divided into the following four categories Z1 .. Z4:

Z1

Sporadic traffic

Z2

Moderate traffic

Z3

heavy traffic

Z4

Traffic jam / "Stop and Go"

For the automated determination of the traffic condition autonomous and stationary solutions are proposed Service.

EP 0 789 341 A1 (and EP 0 789 341 B1) describes an autonomous vehicle Detection of traffic congestion revealed at the individual Vehicles are provided with a device that the Vehicle speed recorded continuously, this Assigns speed to specified speed classes and an integral with the assigned speed classes form over time, its value to form a backlog is used. Together with the location of the vehicle is the storage volume e.g. via GSM to a data collection point transfer.

WO 01/35372 contains an example of a stationary solution Video surveillance system described using so-called Pattern recognition on the license plates of two different cars Digits recorded and by comparing the times a traffic condition is detected and a monitoring point or reports to the police.

With the aforementioned video surveillance, the problem occurs of Reflections that occur at certain sun positions or manifest as very unpleasant when it rains. Furthermore video surveillance requires a separate infrastructure Cameras and corresponding wired or wireless Connections, their provision and maintenance consuming is.

EP 1 056 063 A1 (Siemens Schweiz AG) describes a method for Determination of the traffic condition of a traffic segment discloses a model having a Kalman filter underlying. The procedure is not of a particular kind the measurement of traffic volumes.

Regarding an acoustic evaluation is from the railway technology known to determine the exact source of the noise and the type of spread on a railway line. a so-called microphone array is set up and the analysis of the geographically different usually mostly based on microphones arranged at different heights make a statement about the exact source of noise on a train or locomotive and allowed the path of expansion.

In US 6,195,608 (B1 Lucent Technologies Inc.) for Detection of a moving vehicle within a predetermined one Zone proposed a system that has a plurality of microphones aligned to the specified zone, various Bandpass filter, correlation and integration means having. The detection itself takes place due to the exceeding of the signal emitted by the integration means a certain threshold.

Such a system for the detection of road traffic is with regard to the arrangement along a street, e.g. in the Distance of 2 - 3 km with an unsustainable effort connected.

The present invention is therefore based on the object specify a procedure that allows the state of traffic to automatically detect a section of a road that an already existing infrastructure can be used.

This object is achieved by the specified in claim 1 Measures solved. Advantageous embodiments of the invention are specified in the dependent claims.

The method according to the invention for recording traffic conditions requires on site, i.e. on the street none additional infrastructure because existing microphones e.g. used by the emergency pillars can be. In particular, no microphone arrays needed, and the inventive monophonic process can easily be provided on a computer with any additional digital signal processors are implemented and in a traffic control and / or traffic information system be involved.

i) Dadurch, dass
im Verfahrensschritt C für drei im hörbaren Bereich liegende Frequenzbänder ein zweites Merkmal (M2) generiert wird, das aus dem Quotienten der mittleren Leistung im mittleren Band und der mittleren Leistung im unteren Band gebildet wird;
erfordert die Generierung des zweiten Merkmals keine grosse Rechenkapazität (Patentanspruch 6).

ii) Dadurch, dass
die Verkehrszustände geordnet sind und dass aus den im Verfahrensschritt B generierten Merkmalen zu einem unmittelbar vorangehend erzeugten Verkehrszustandsindikator im Verfahrensschritt D nur ein benachbarter Verkehrszustandsindikator erzeugt wird; ist das Verfahren gegenüber akustischen Störeinflüssen robust (Patentanspruch 8).

The following additional advantages can result:

i) In that
a second feature (M2) is generated in method step C for three frequency bands lying in the audible range, which is formed from the quotient of the mean power in the middle band and the mean power in the lower band;
the generation of the second feature does not require a large computing capacity (claim 6).

ii) In that
the traffic conditions are ordered and that only a neighboring traffic condition indicator is generated from the features generated in method step B for a traffic state indicator generated immediately before in method step D; the method is robust against acoustic interference (claim 8).

Figur 1

Disposition an einer richtungsgetrennten Strasse mit einer Notrufsäule;

Figur 2

Vergrösserte Aufsicht einer Notrufsäule in der Disposition gemäss Figur 1;

Figur 3

Ablauf des erfindungsgemässen Verfahrens in der Uebersicht;

Figur 4

Funktionseinheiten zur Vorverarbeitung;

Figur 5

Signalfluss für die Merkmalsextraktion;

Figur 6

Zustandsdiagramm mit zulässigen Zustandsübergängen.

The invention is explained in more detail below with reference to the drawing, for example. Show:

Figure 1

Disposition on a direction-separated street with an emergency call pillar;

Figure 2

Enlarged supervision of an emergency call pillar in the disposition according to FIG. 1;

Figure 3

Overview of the process according to the invention;

Figure 4

Functional units for preprocessing;

Figure 5

Signal flow for feature extraction;

Figure 6

State diagram with permissible state transitions.

The description of an embodiment of the present invention is based on the four traffic conditions mentioned at the beginning, which are quantified as follows by way of example (the abbreviation "Fz" stands for vehicle): symbol description Description, quantification Z1 Sporadic traffic Flow J with J <15 [Fz / min] Z2 Moderate traffic 15 <J <35 [vehicle / min] Z3 heavy traffic J> 35 [vehicle / min] Z4 Traffic jam / "Stop and Go" Density D high [Fz / 100 m] and v <15 [km / h]

The used by the inventive method and in all The infrastructure that is already present is shown in FIG. 1. Along a street are at intervals of approx. 500 m up to about 2 km of emergency telephones arranged one by one Head office such as microphone that can be activated by a police station exhibit. This control center has an evaluation unit that based on predefined settings or as required by intervention of an official contained in the emergency pillars Microphones activated. The disposition of such emergency pillars is mostly individual, i.e. the distance and the relative position to the at least one lane is dependent local conditions and, depending on the country, legal requirements Requirements. In order not to endanger people provoke, is usually also on the opposite side an emergency call column is arranged in the same position (not shown in Fig. 1). In the disposition according to Figure 1 the microphone is arranged in the direction of travel r '' is possible also a disposition of the reception cone orthogonal to the lane 2 ', 2' 'or against the direction of travel r'.

The microphones of the individual emergency call pillars can be of the mentioned Central can be activated or deactivated. For the The inventive method is in one embodiment provided that a central evaluation unit the different Emergency telephones and the microphones contained therein activated and during a previously configured, i.e. adjustable interval during an also adjustable Duration activated and absorbs the ambient noise and as a so-called noise pattern to the central evaluation unit transmitted. It goes without saying that the use of the Emergency column not affected by a person in an emergency. In such a case, the transfer of the Noise pattern canceled and made at a later date then another activation of the microphone.

3 shows an overview of the various process steps of the inventive method with the in a Evaluation unit implemented blocks 31, 32, 33 and 34.

The ambient noise is contained in the emergency column 3 Microphone 6 added. This microphone 6 is through the evaluation unit mentioned is activated, for example the ambient noise is periodically during an interval of e.g. 60 s recorded by the microphone and as noise pattern A1 (also called frame) transmitted to the evaluation unit. The For example, periodicity is 15 minutes. In block 31 As part of the evaluation unit, a noise pattern A1 with a Frequency of e.g. 8 kHz sampled, and thereby samples A1 'generated, which represent individual sections of 125 µs duration. These samples are preferably digital Form here with a resolution of e.g. 16 bit.

F1

Tiefpass : bis 200 Hz;

F2

Bandpass : 200 - 2000 Hz;

F3

Hochpass : 2000 - 4000 Hz.

In block 32, the sample values are preprocessed, for example the 60x8000 digital sample values A1 'described above, which represent a noise pattern A1 of 1 minute duration. The detailed method steps in block 32 are shown in FIG. 4 with the method steps or units 411 .. 434. The sample values A1 'are each fed to a low pass 411, a band pass 421 and a high pass 431, the frequencies of which follow the disjoint bands F1, F2 and F3 include:

F1

Low pass: up to 200 Hz;

F2

Band pass: 200 - 2000 Hz;

F3

High pass: 2000 - 4000 Hz.

The digital audio signals of the aforementioned three bands are then subjected to smoothing in units 413, 423 and 433, and undersampling in units 414, 424 and 434 results in a reduction in the number of samples without loss of information , this last statement is based on the Shannon-Whittaker theorem. The signals generated by the units 414, 424 and 434 are designated S1, S2 and S3, the signal S1 standing for the frequency band up to 200 Hz and S2 and S3 correspondingly in the order for the other two frequency bands mentioned above. A signal Sx (x = 1, 2, 3) comprises, for example, 60x8 to 60x16 samples, which represent a noise pattern A1 of 1 minute duration in the relevant frequency band. The number 60 refers to the expansion from 1 s to one minute. The signals S1, S2, S3 themselves represent a level which corresponds to the sound pressure, which, apart from a scaling, has the unit [N / m ² ].

In block 33, a feature extraction takes place, which is shown in more detail in FIG. 5 with the units 51, 52, 53. In unit 51, the ripple of the power is determined from signal S2 by differentiation according to time t, the amount being derived from the derivative. It would also be possible to square the derivative instead of the amount. The ripple determined in this way can be used as a feature M1 for the traffic conditions Z1 "moderate traffic" and Z3 "heavy traffic" to distinguish the condition Z4 "traffic jam". The signals S1 and S2 are subjected to a spectral performance comparison by the unit 52: based on extensive measurements, it has been shown that the quotient listed below is somewhat constant in the traffic condition Z4 traffic jam and can therefore be used as a feature M2 for the traffic condition Z4: average power P in the band 200 - 2000 Hz average power P in the band up to 200 Hz

Also due to measurements at a tunnel entrance It has been shown that the value of this quotient is resistant in one is against weather influences such as Rain and although as long as the noise of the rain a certain proportion of the noise emitted by the vehicles.

In block 53, a power peak is determined from signal S3. The characteristic M3 obtained in this way is suitable for the Determination of the traffic conditions Z1, Z2 and Z3.

{Z1, Z2, Z3, Z4}.
Z(t) wird als Zustandsindikator bezeichnet und ordnet den im Block 33 ermittelten Merkmalstripeln {M1, M2,M3} einen Strassenverkehrszustand aus dem Bildbereich Z1, Z2, Z3 Z4 zu. Diese Abbildung erfolgt aufgrund vorgängig gespeicherter Merkmalswerte und Merkmalswertebereiche aus dem jeweiligen Wertebereich von M1, M2 und M3. Da der Strassenverkehrszustand zwar diskret jedoch in der zeitlichen Abfolge geordnet bzw. stetig ist, beinhaltet die vorgenannte Abbildung Z(t) eine Zustandsmaschine (auch Zustandsautomat genannt) gemäss der Fig. 6. Dadurch wird erreicht, dass der Zustandsindikator zeitlich hintereinander nur im Sinne der Ordnung benachbarter Zustände Z(t) und Z(t+Δt) generiert. In diesem Kontext muss die vorgenannte Abbildung nach einer erfolgten Initialisierung gesehen werden als:

The features defined above are supplied to the unit 34 - see FIG. 3 in this regard - for the illustration Z (t): {M1, M2, M3}

{Z1, Z2, Z3, Z4}.
Z (t) is referred to as the status indicator and assigns the feature triples {M1, M2, M3} determined in block 33 to a road traffic condition from the image area Z1, Z2, Z3 Z4. This mapping takes place on the basis of previously stored characteristic values and characteristic value ranges from the respective value range of M1, M2 and M3. Since the road traffic state is discrete in order of time or continuous, the aforementioned figure Z (t) contains a state machine (also called a state machine) according to FIG. 6. This ensures that the state indicator is only one after the other in the sense of Order of neighboring states Z (t) and Z (t + Δt) is generated. In this context, after the initialization, the above illustration must be seen as:

Robustness can be achieved with such a state machine significantly increase the inventive method in which Outliers the actual traffic condition is not wrong can play.

There are several emergency telephones along a street arranged, the robustness can be further increased be, at most with a time delay the determined traffic conditions are analyzed is on the one hand the case of "collision" when drifting apart of traffic conditions between two immediately adjacent ones Detection points can be determined directly. on the other hand is the temporal in particular with the states Z1 and Z2 Delay between the two detection points mentioned can be analyzed and therefore gives an even better one Quality of the recording of traffic conditions.

Above is a central embodiment, i.e. implementation of the method according to the invention have been described, where the given values are to be regarded as examples, which in particular of the type of transmission and thus of the bandwidth between the emergency column and the microphone are. To be independent of this bandwidth or its limitation The method according to the invention is also to be Can be implemented decentrally, that is, each emergency call point points directly an evaluation unit of the aforementioned type on and transmits the traffic condition indicator Z (t) to one superordinate surveillance body.

The type of feature extraction is neither indicated on the three Frequency bands F1, F2 and F3 still on the specified Ripple determination, the spectral performance comparison through quotient formation or the determination of peak performance limited.

List of the reference symbols used

1

central reservation

2 ', 2' '

lane

3

Notrufsäule

4

Access area to the emergency column 3

5

Moving vehicles

6

Microphone embedded in the emergency column 6

7

turn signal

31

Signal acquisition and signal conversion

32

preprocessing

33

feature extraction

34

Classification of characteristics and assignment to traffic conditions

411

lowpass

421

bandpass

431

highpass

412, 422, 432

power determination

413, 423, 433

Smoothing

414, 424, 434

Sample reduction

51

Determination of the ripple of the power

52

Spectral performance comparison

53

Peak power determination

A1

Noise pattern, frame, audio signal analog

A1 '

60x8000 samples from A1, audio signal digital

d

current path of sound to microphone 6

F1, F2, F3

Disjoint and ordered frequency bands

M1, M2, M3

Features extracted from signals S1, S2, S3

r ', r "

provided for the lanes 2 ', 2' ' driving directions

S1, S2, S3

Signal in different frequency bands

Z (t)

Traffic status indicator

Z1, Z2, Z3, Z4

Traffic conditions, expressions of Z (t)

Claims (8)

1. A method for determining road traffic conditions, wherein ambient noise is recorded by a microphone (6) disposed on a road (2', 2'') and transmitted as a noise sample (A1) to an evaluation unit,
   characterised by the method steps:

   A in the evaluation unit, a plurality of sample values (A1') is generated from the noise sample (A1) (31);

   B from the sample values (A1') generated in method step A, signals (S1, S2, S3) are generated which represent a level made up of different frequency bands (F1, F2, F3) (32);

   C from the signals (S1, S2, S3) generated in method step B, features (M1, M2, M3) are generated by means of an analysis in the frequency or time domain (33);

   D from the features (M1, M2, M3) generated in method step C, a traffic condition indicator ( Z(t) ) is generated on the basis of a stored assignment of feature value ranges to traffic conditions (Z1, Z2, Z3, Z4) (34).
2. The method according to Claim 1,
   characterised in that
   the noise sample (A1) is recorded from the ambient noise by the microphone (6) in an adjustable time interval during an adjustable time period.
3. The method according to Claim 1 or 2,
   characterised in that
   in method step B, a first, second and third signal (S1, S2, S3) are generated, each of which originates from a specific frequency band of the noise sample (A1).
4. The method according to Claim 3,
   characterised in that
   a first feature (M1) is generated which represents the ripple of the second signal (S2) generated in method step B.
5. The method according to Claim 3 or 4,
   characterised in that
   in method step B the signals (S1, S2, S3) represent the power.
6. The method according to Claim 5,
   characterised in that
   in method step C a second feature (M2) is generated for three frequency bands (F1, F2, F3) lying in the audible range, said feature being formed from the quotient of the mean power in the middle band (F2) and the mean power in the lower band (F1).
7. The method according to Claim 5 or 6,
   characterised in that
   in method step C a third feature (M3) is generated which represents power peaks.
8. The method according to one of Claims 1 to 7,
   characterised in that
   the traffic conditions (Z1, Z2, Z3, Z4) are graded and that in method step D, from the features (M1, M2, M3) generated in method step C relating to a traffic condition indicator ( Z(t) ) generated immediately beforehand, only an adjacent traffic condition indicator ( Z(t+Δt) ) is generated.

Images