CA 02826723 2013-09-10
Method and Apparatus For Controlling Surveillance System With Gesture And/Or
Audio Commands
Field
This disclosure relates generally to controlling a surveillance system with
gesture and/or
audio commands.
Background
Controlling various functionalities of a digital surveillance system using
standard input
devices like a keyboard, joystick and/or mouse can be restrictive and slow. An
operator
needs to physically manipulate the standard input device to use the various
functionalities of the system and to be able to view surveillance data. The
learning
curve for using a surveillance system using such input devices can be steep.
Further,
operation of the system is usually restricted to one operator at a time, who
is in direct
arms' reach of the input device connected to the system.
Summary Of Invention
One objective is to provide an improved method for controlling certain
operations of a
surveillance system; such improved method can, for example, provide a faster,
easier
and more intuitive method for controlling video management software of a
surveillance
camera.
According to one aspect of the invention, there is provided a computer-
implemented
method which controls aspects of a surveillance system using gestures and/or
voice
commands. The method comprises: receiving one or both of an operator's
skeleton
input data and voice input data from a gesture detection device and a
microphone;
matching one or both of the received skeleton input data with a gesture stored
in a
database and the received voice input data with a text string stored in the
database;
matching one or both of the gesture and text string to a corresponding video
management program command stored on the database; and transmitting the one or
more video management program commands to a video management program of the
1
CA 02826723 2013-09-10
surveillance system.
The step of matching the receiving skeleton data with a gesture can comprise
storing
frames of skeleton input data received over a defined time frame in a gesture
list, then
determining whether the stored frames match a set of parameters that define
the
gesture. The database can comprise multiple gestures comprising discrete
physical
gestures and continuous physical gestures. In such case, the method further
comprises determining an operating state of the video management program, then
matching the received skeleton data with a gesture selected from a group of
discrete
physical gestures or a group of continuous physical gestures depending on the
determined operating mode. Examples of discrete physical gestures include
"swipe left",
"swipe right", "swipe up", and "swipe down". The corresponding video
management
program command to the "swipe left" gesture causes a selected panel in a grid
of
panels displayed by the video management program to switch places a panel to
the left.
The corresponding video management program command to the "swipe right"
gesture
causes a selected panel in a grid of panels displayed by the video management
program to switch places a panel to the right. The "swipe up" gesture causes a
selected
panel in a grid of panels displayed by the video management program to switch
places
a panel above. The "swipe down" gesture causes a selected panel in a grid of
panels
displayed by the video management program to switch places a panel below.
Examples of continuous physical gestures include "TimeLine Scrubbing" and
"Video
Panning and Zooming". The corresponding video management program command to
the "TimeLine Scrubbing" gesture causes a cursor of a displayed timeline to be
moved
in the direction of the TimeLine Scrubbing gesture. The corresponding video
management program command to the "Video Panning and Zooming" gesture causes a
part of a video stream to be moved about a panel in the direction of the Video
Panning
and Zooming gesture.
The step of matching the received voice input data with a text string stored
in the
database can comprise using a speech recognition engine to convert the voice
input
data into a text string, then determining if the converted text string matches
a
recognized text string stored on the database.
2
CA 02826723 2013-09-10
According to another aspect of the invention, there is provided a surveillance
system
comprising: a surveillance camera; one or both of a gesture capture device and
a
microphone; and a computer communicative with the surveillance camera and with
one
or both of the gesture capture device and the microphone. The computer
comprises a
processor and a computer readable medium having stored thereon a video
management program and an interpreting software component program. The
interpreting software component program comprises a database and program code
executable by the processor to perform a method comprising the following
steps:
receiving one or both of an operator's skeleton input data from the gesture
capture
device and voice input data from the microphone; matching one or both of the
received
skeleton input data with a gesture stored in the database and the received
voice input
data with a text string stored in the database, using the processor; matching
one or both
of the gesture and text string to a corresponding video management program
command
stored on the database, using the processor; and transmitting the one or more
video
management program commands to the video management program.
According to yet another aspect of the invention, there is provided a computer
readable
medium having stored thereon an interpreting software program comprising a
database
and program code executable by a processor to perform a method for controlling
aspects of a surveillance system using gestures or voice commands. This method
comprises: receiving one or both of an operator's skeleton input data and
voice input
data; using a processor to match one or both of the received skeleton input
data with a
gesture stored on the database and the received voice input data with a text
string
stored on the database; using the processor to match one or both of the
gesture and
text string to a corresponding video management program command stored on the
database; and transmitting the one or more video management program commands
to
a video management program of a surveillance system.
Brief Description of Drawings
Figure 1 is a schematic block diagram of components of a surveillance system
according to one embodiment.
3
CA 02826723 2013-09-10
Figure 2 is a logic diagram of the functions performed by components of the
surveillance system including an interpreting software component.
Figure 3 is a flowchart of steps performed by the interpreting software
component in
response to gesture and audio inputs received from the input device shown in
Figure 2.
Detailed Description
Embodiments described herein relate to a computer implemented method for
controlling aspects of a surveillance system using gestures captured by a
gesture
detection device (e.g. one or more cameras) and audio commands captured by a
microphone. More particularly, the described embodiments comprise an
interpreting
software component that is communicative with a video management program of
the
system, to send command instructions to the video management program in place
of
manual input devices like a keyboard, joystick and mouse. The interpreting
software
component is also communicative with a gesture detection device and a
microphone to
receive operator skeleton input data and operator voice input data. The
interpreting
software component comprises a database of command instructions compatible
with
the video management program and gestures and/or text strings associated with
the
command instructions. The interpreting software component also comprises a set
of
gesture interpretation algorithms for associating received skeleton input data
with a
gesture in the database, and speech recognition algorithms for associating
received
voice input data with a text string in the database. The speech recognition
algorithms
can be provided by an application program interface (API) integrated into or
communicative with the interpreting software component. Once the gestures
and/or
text strings have been determined by the respective gesture interpretation and
speech
recognition algorithms, the interpreting software component then associates a
command instruction in the database with the gestures and/or text strings, and
sends
the command instructions to the video management program.
Referring to Figure 1 and according to one embodiment, a surveillance system 1
generally comprises a surveillance camera 10, a computer 11 communicative with
the
camera 10, and a gesture and voice input device 12 communicative with the
computer
4
CA 02826723 2013-09-10
11. The surveillance camera 10 has a housing 30 for housing camera components,
and
a movable mount 32 for mounting the rest of the camera 10 to a surface such as
a
ceiling. A zoom lens 14 is mounted at the front of the housing 30. A camera
communications cable 16 is coupled to the camera 10 and the computer 11 and
enables two-way communication between the camera 10 and computer 11 such that
image data captured by the camera 10 can be transmitted to the computer 11 and
camera operation commands can be transmitted by the computer 11 to the camera
10.
Suitable cameras for use in the system 1 include those provided by Avigilon
Corporation, such as their HD Bullet Cameras.
A user interface communications cable 18 is coupled to the computer 11 and
gesture
and voice input device 12 and enables two-way communication between the
gesture
and voice input device 12 and the computer 11 such that a user's skeleton
input data
and voice input data captured by the gesture and voice input device 12 can be
transmitted to the computer 11, and device operating commands can be
transmitted
from the computer 11 to the gesture and voice input device 12.
In this embodiment, a single device 12 contains both gesture and voice
detection
sensors. Alternatively, two separate devices can be provided to record gesture
and
voice inputs separately (not shown). One suitable gesture and voice input
device 12 is
the Microsoft's KinectTM sensor array and related software. The KinectTM senor
array is
a physical device that contains cameras, a microphone array and an
accelerometer, as
well as a software pipeline that processes color, depth, and skeleton data.
The related
software includes Kinect for WindowsTM API ("Kinect API"), which comprises a
Natural
User Interface (NUI) that allows an applications software developer to access
the audio,
color image, and depth data streamed from the Kinect sensor array for a Kinect-
enabled
application. The Kinect API includes algorithms that can recognize and track a
human
body, by converting depth data into skeleton joints in the human body;
skeletons for up
to two people at a time can be created and tracked. The Kinect API also
integrates with
the Microsoft SpeechTM API to allow a developer to implement a speech
recognition
engine into the Kinect-enabled application.
5
CA 02826723 2013-09-10
Referring now to Figures 1 and 2, the computer 11 comprises a display 20, a
processor
(not shown) and a computer readable medium 23 having stored thereon a video
management program 22 and the interpreting software component 24. The
processor,
display and memory can be part of a personal computer like a laptop or desktop
PC, or
be components of a commercial scale server and client system in a manner that
is well
known in the art. Both the video management program 22 and interpreting
software
component 24 are executable by the processor to implement method steps encoded
in
the respective software programs. In particular, the interpreting software
component 24
will receive operator skeleton and voice input data from the gesture and voice
input
device 12, associate gestures and/or text strings stored in the computer
storage
medium with each received skeleton and voice input data, associate the
associated
gestures and/or text strings with a corresponding video management program
command, then transmit the video management program command(s) to the video
management program 22.
The video management program 22 can be an open source software application
such
as ZoneMinderTm, or be a proprietary software application such as the Avigilon
Action
CentreTM. Such programs typically support cameras from a number of
manufacturers,
and can be integrated into legacy third party systems. The video management
program
22 should be communicative with one or more surveillance cameras to receive
captured
image data (in the form of a series of image stills and/or a continuous video
stream) and
be operable to control operation of the surveillance camera(s), as well as to
record,
display and manipulate images and videos taken by the camera(s).
For example, the Avigilon Action CentreTm program can interface with up to 128
cameras per server, provides for joystick control and mobile device input
control, and
has a number of recording, search and playback features including: jog dial
search,
thumbnail image search, pixel search (to detect motion events within a scene).
The
Avigilon Action CentreTM can also display multiple video streams concurrently
on a
single display, change the layout of the video streams on the display, call up
a video
stream from a camera by the camera's assigned name and number, change zoom
levels of video streams, and switch between live and recorded video.
6
CA 02826723 2013-09-10
As another example, the ZoneMinderTm program has many of the same features as
the
Avigilon Action CetnreTM, including:
= Multiple zones (regions of interest) can be defined per camera. Each can
have a
different sensitivity or be ignored altogether.
. Web
interface allowing full control of system or cameras as well as live views and
event replays.
= Supports live video in MPEG video, multi-part JPEG, and stills formats.
= Supports event replay in MPEG video, multi-part JPEG, and stills formats
along
with statistics detail.
= User defined filters allowing selection of any number of events by
combination of
characteristics in any order.
= Event notification by e-mail or SMS, including attached still images or
video of
specific events by filter.
= Automatic uploading of matching events to external FTP storage for
archiving
and data security.
= Includes bi-directional X10 (home automation protocol) integration
allowing X10
signals to control when video is captured and for motion detection to trigger
X10
devices.
= Partitioned design allows other hardware interfacing protocols to be
added for
support of alarm panels, etc.
The interpreting software component 24 can be adapted to interface with the
specific
functions of each video management program. For example, with the ZoneMinder
program, the interpreting software component 24 will interface with the API of
the
ZoneMinderTm program relating to controlling various operations of the
program, such
as control of zooming and panning values. The interpreting software component
24
thus uses the API to send commands in the form of X, Y, and Z values needed
for such
panning and zooming, or any other form the API requires. ZoneMinderTm uses a
web
interface to control a camera; conventional interfacing with ZoneMinderTm
involves
sending commands via a mouse or keyboard, which are interpreted using a web
technology such as javascript or HTML and then sent to the ZoneMinderTm server
or
other components. In a similar manner, the interpreting software component 24
can be
7
CA 02826723 2013-09-10
programmed to generate commands in the form of appropriate javascript or HTML
values from gesture and text strings, which have been interpreted from
skeleton and
voice input data (as will be explained in detail below).
Instead of interfacing with a video management program's API for external
input control,
the interpreting software component 24 can interface directly with the video
management program in which the program code of the video management program
would be modified as necessary.
Voice-Activated Commands
The interpreting software component 24 in this embodiment is a Kinect-enabled
application that incorporates the KinectTM API and the Microsoft SpeechTM API.
Alternatively, the interpreting software component 24 can be adapted for use
with other
gesture and voice input sensor devices, and would be modified to use the
software
interface APIs associated with those other devices in the manner as is known
in the art.
The Microsoft SpeechTM API ("SAPI") is speech recognition engine. The SAPI
implements the low-level details need to control and manage the real-time
operations of
a speech recognition engine. In particular, the SAPI converts the voice input
data into
computer readable text strings or files in a manner that is known in the art.
The
interpreting software component 24 uses the SAPI to convert the voice input
data into a
text string, and then tries to match the text string with one of the video
management
program commands stored in memory. The following is a non-exhaustive and
exemplary list of possible video management program commands and their
associated
text strings converted from voice input data:
= Start tracking ¨ text string: "Start Tracking".
= Stop tracking ¨ text string: "Stop Tracking".
= Maximize a selected video panel in a grid of video panels to cover the
entire
monitor ¨ text string: "Maximize".
= Minimize a video panel covering the entire screen into one panel amongst
a grid
of video panels ¨ text string: "Minimize".
8
CA 02826723 2013-09-10
The following 3 commands are used when the video management program 22 shows a
grid of video panels on the display, with the panels numbered consecutively
left to right,
row by row, starting at 1 for top most left panel and ending at the number of
panels for
the bottom most right panel. For example, in a 3 rows by 4 columns grid, the
top most
left panel would receive the panel number 1, the second panel on the second
row would
be numbered panel 6 and the bottom most right panel would be number 12:
= For each video panel in a grid of video panels, display the panel number
¨ text
string: "Show Panel Numbers".
= Display a video stream from a specific named camera (cameras can be
assigned
names in a digital surveillance system) onto a numbered video panel in a grid
of
video panels ¨ text string: "Add camera <camera name> to panel <panel
number>". For example, "Add camera Elevator to panel five.
= Focus a specific video panel by its number in a grid of video panels (a
focused
panel will be later the target of other actions like zooming) ¨ text string:
"Select
Panel <number>". For example, "Select Panel three".
= Display recorded video data instead of live video data - text string:
"Recorded
Video".
= Display live video data instead of recorded video data - text string:
"Live Video".
= Select a specific day from which to display recorded video data - text
string: "Go
To <Month> <day of month>. For example "Go To December Eight".
= Select a specific time of day from the selected day from which to display
recorded video data - text string: "Set Time To <time of day> <AM/PM>. For
example "Set Time To 4 twelve PM".
= Set the focus of the digital surveillance system to its "TimeLine" (the
TimeLine is
a component which shows a time range and can be used to pick a specific date
and time from which to show recorded video data) ¨ text string: "Select Time
Line".
= Narrow the time range displayed in the Time Line (which enables a finer
resolution control of time values selected in the TimeLine). This command will
execute only if the focus of the digital surveillance system is on the
TimeLine
9
CA 02826723 2013-09-10
(see the previous command). This action can be done in continuous steps, e.g.
one every 0.1 seconds, each narrowing the displayed time range in the TimeLine
by 5% and will continue until stopped (see the Stop audio command) or until
the
highest level of time resolution has been reached - text string: "Zoom In".
= Increase the time range displayed in the Time Line. This command is the
identical to the "Zoom In" audio command in all aspects except it will
increase the
time range displayed in the TimeLine rather than narrow it - text string:
"Zoom
Out".
= Stop narrowing or increasing the time range in the TimeLine. This will
have effect
only if the Zoom In or Out command was given before it - text string: ¨
"Stop".
Gesture-Activated Commands
In this embodiment, the Kinect sensor array 12 will convert captured video of
an
operator 26 and send skeleton input data via the Kinect API to the
interpreting software
component 24. The interpreting software component 24 contains a set of gesture
interpretation algorithms which can determine whether the skeleton input data
captured
over a defined time frame (e.g. 0.25 seconds) matches certain parameters that
define a
recognized gesture. As frames of skeleton input data are inputted from the
gesture and
voice input device 12 over the defined time frame, a gesture list is built
containing joint
locations and times for each location (for example, a gesture list can be set
to contain
the latest 35 locations and times) and stored on a database of the computer
memory.
This gesture list is later used to determine if the recorded locations over
time match a
recognized gesture in the database.
The recognized gestures can be categorized generally as discrete physical
gestures
and continuous physical gestures. Discrete physical gestures are those that
occur only
once during a defined time frame, i.e. they are completed within the time
frame and do
not continue continuously beyond the end of the time frame. Continuous
physical
gestures continue beyond the defined time frame, and do not stop until the
operator
lowers his/her hand or tells the system 1 to stop tracking. The interpreting
software
component will access a set of discrete physical gestures or a set of
continuous
CA 02826723 2013-09-10
,
physical gestures stored on the database based on which operating state the
video
management program is in at the time the operator made the gesture. That is,
the
gesture interpreting algorithm will try to match the skeleton input data with
either a
discrete physical gesture or with a continuous physical gesture depending on
the
operating state of the video management program.
The recognized discrete physical gestures include "Swipe Left", "Swipe Right",
"Swipe
Up", and "Swipe Down", and the recognized continuous physical gestures include
"TimeLine Scrubbing" and "Video Panning and Zooming". Each of these recognized
gestures are stored in the database along with an associated video management
program command. The Swipe Left gesture is triggered by a swipe left of the
right hand
and is associated with a command that causes a selected panel in a grid of
panels to
switch places with the panel on its left. The Swipe Right gesture is triggered
by a swipe
right of the operator's right hand and is associated with a command that
causes a
selected panel in a grid of panels to switch places with the panel on its
right. The Swipe
Up gesture is triggered by a swipe up of the right hand and is associated with
a
command that causes a selected panel in a grid of panels to switch places with
the
panel above it. The Swipe Down gesture is triggered by a swipe down of the
right hand
and is associated with a command that causes a selected panel in a grid of
panels to
switch places with the below it. The TimeLine Scrubbing gesture is triggered
when a
Select Time Line state of the video management program is activated that
focuses on a
time line, by moving the right hand left or right and is associated with a
command that
moves a cursor of the displayed time line in the direction of the hand
movement; this
enables control of which recorded display data is displayed. The Video Panning
and
Zooming gesture is triggered when a Select Panel state of the video management
program 22 has been activated, by moving the right hand in space (left, right,
up, down,
towards and away) relative to the gesture and voice input device 12 and is
associated
with selecting a particular part of the video stream to display on the panel.
A gesture interpretation algorithm for interpreting swipes is provided to
determine if a
swipe was gestured, based on the following parameters:
11
CA 02826723 2013-09-10
1. The hand must be at least a certain distance (e.g. 0.4 meters) away from
the
operator's chest.
2. The swipe had to be completed with the defined time frame (e.g. 0.25
seconds).
3. Get all locations of the joint (e.g. the right hand) within the last
defined time frame
(0.25 seconds) from the list.
4. There has to be at least a threshold amount of joint locations (e.g. 7)
which
occurred in the last defined time frame in the list.
5. All the locations of the joint in the last time frame need to be following
the
direction of the swipe (right or left or up or down). For that to happen, the
joints
locations over the time frame need to advance relative to each other in the
same
direction (the direction of the swipe) without going at the opposite direction
at all
and without deviating to a perpendicular direction more than a threshold (e.g.
0.1
meters).
6. If the distance between the latest location of the joint and the oldest
location of
the joint within the defined time frame is over a threshold (e.g. 0.15 meter)
a
swipe is identified.
The following is an example of the gesture interpretation algorithm
determining whether
received skeleton input data indicates the "Swipe Left" gesture:
1. Get the latest location of the right hand and the chest of the operator.
2. Add the latest location to the gesture list of latest locations.
3. Check the distance between the right hand and chest. If it is more than 0.4
meters, then conclude there was no swipe and stop the gesture interpretation
algorithm.
4. Get all locations of the joint from the gesture list that occurred within
the last 0.25
seconds and store in a separate list entitled Time Frame List.
5. If there are less than seven (7) locations in the Time Frame List, then
conclude
that there was no swipe and stop the gesture interpretation algorithm.
6. Compare the latest joint location with all other joint location in the Time
Frame
List, and confirm that the latest joint location is to the left of all other
locations in
the list.
12
CA 02826723 2013-09-10
7. Compare the latest joint location with all other joint location in the Time
Frame
List. It must have a vertical distance (up or down) from any of them no higher
than 0.1 meters.
8. If the horizontal distance between the first (oldest) member of the Time
Frame
List and the latest (just arrived) member is greater than 0.15 meter then
register
a "Swipe Left" gesture.
Referring now to Figure 3, the interpreting software component 24 contains a
number of
programmed method steps that are executed by the computer processor to
interpret the
input data received from the user interface device 12 into commands that are
compatible with the video management program 22, then output those commands to
the
video management program 22.
At start-up the processor loads both the video management program 22 and the
interpreting software component 24 (step 50). As part of the loading process,
a list of
gestures and text strings and the corresponding commands are loaded into
memory of
the processor (step 52), and a communications connection is established with
the
gesture and voice input device 12 and starts "listening" for input data (step
54). When
new input data arrives from the gesture and voice input device 12 (step 56),
the
interpreting software component 24 determines whether the new data is skeleton
input
data or voice input data (step 58).
When the new data is determined to be skeleton input data, the interpreting
software
component 24 updates the gesture list over the defined time frame with the new
skeleton input data (step 60). The interpreting software component 24 then
determines
if the video management program 22 is in an operating state that is commanded
by
discrete physical gestures or by continuous physical gestures (step 62). If
the latter,
then the interpreting software component 24 executes the gesture
interpretation
algorithm for continuous physical gestures to match the skeleton input data
stored on
the gesture list to a recognized continuous physical gesture, then associates
a particular
video management program command with the matched continuous physical gesture,
e.g. a particular panning/zooming command if the video management program is
in a
13
CA 02826723 2013-09-10
selected panel mode or a particular timeline scrubbing command if the video
management program is in timeline focused mode (step 64). The interpreting
software
component then sends the associated command to the video management program
(step 65); for example, video panning and zooming data is sent in the form of
X, Y, Z
values to the video management program 22.
When the video management program is in an operating state that is commanded
by
discrete physical gestures, the interpreting software component applies the
gesture
interpretation algorithm to first determine if a completed gesture was
received (step 66),
by matching the skeleton input data in the gesture list to a recognized
discrete physical
gesture. If the gesture list does not contain a complete discrete physical
gesture, then
the interpreting software component 24 returns to step 56 to receive new input
data
(step 68); if the gesture list does contain a complete discrete physical
gesture, then the
matched gesture is associated with a video management program command (step
69),
and this command is sent to the video management program (Step 72).
When the new input data is determined to be voice input data, then the
interpreting
software component 24 executes the Microsoft Speech API ("SAPI") speech
recognition
engine to convert the voice input data into a text string, and then matches
the text string
to a recognized text string in the database (step 74). If a match is found,
then the video
management program command associated with the matched text string is sent to
the
video management program (Step 76). If a match was not found, the interpreting
software component waits for more input data (Step 70) and repeats the above
process.
While particular embodiments have been described in the foregoing, it is to be
understood that other embodiments are possible and are intended to be included
herein. It will be clear to any person skilled in the art that modifications
of and
adjustments to the foregoing embodiments, not shown, are possible. The scope
of the
claims should not be limited by the preferred embodiments set forth in the
examples,
but should be given the broadest interpretation consistent with the
description as a
whole.
Example
14
CA 02826723 2013-09-10
The following is exemplary psedo-code for a gesture interpretation algorithm
for
matching skeleton input data to a video panning and zooming gesture:
To avoid jitteriness, the algorithm uses only every second frame of data from
the
gesture input device 12.
X,Y and Z represent a point in the 3D space in front of the gesture sensor.
For each frame of data arriving from the gesture sensor:
If the distance of the right hand to the center of the shoulders is less than
0.3, ignore
this frame of data as the hand is too close to the body and the user probably
doesn't
want to be tracked.
If not less than 0.3:
[Pseudo code for panning right and left:]
HorizontaValue = (X value of right hand ¨ X value of right shoulder) divided
by 0.3
IF HorizontaValue > 1 then HorizontaValue = 1
IF HorizontaValue <- 1 then HorizontaValue = -1
HorizontaValue = HorizontaValue + 1
HorizontaValue = HorizontaValue divided by 2
[Pseudo code for panning up and down:]
VerticalValue = (Y value of right shoulder ¨ Y value of right hand) divided by
0.3
IF VerticalValue > 1 then VerticalValue = 1
IF VerticalValue <- 1 then VerticalValue = -1
VerticalValue = VerticalValue + 1
VerticalValue = VerticalValue divided by 2
[Pseudo code for getting the zoom level:]
3DDistanceBetweenRightHandAndRightShoulder =
Square root of (Square value of (X value of right hand ¨ X value of right
shoulder)
Square value of (Y value of right hand ¨ Y value of right shoulder)
Square value of (Z value of right hand ¨ Z value of right shoulder))
If 3DDistanceBetweenRightHandAndRightShoulder > than 0.5
Then 3DDistanceBetweenRightHandAndRightShoulder = 0.5
The variable HandLogicalLocation has an initial value of 0.3 for the first
frame and
afterwards has a start value of its value from the previous data frame
CA 02826723 2013-09-10
HandLogicalLocation = HandLogicalLocation +
(3DDistanceBetweenRightHandAndRightHhoulder - HandLogicalLocation ) divided by
3
3DDistanceBetweenRightHandAndRightHhoulder = HandLogicalLocation
If 3DDistanceBetweenRightHandAndRightHhoulder > 0.5 then
3DDistanceBetweenRightHandAndRightHhoulder = 0.5
If 3DDistanceBetweenRightHandAndRightHhoulder > 0.3 then
3DDistanceBetweenRightHandAndRightHhoulder = 0.3
ZoomValue = (3DDistanceBetweenRightHandAndRightHhoulder ¨ 0.3) divided by
(0.5 ¨ 0.3)
ZoomValue, HorizontaValue and VerticalValue are sent to the digital
surveillance
system and are used to determine which part of the video stream to display.
=
16
