FHWS/IPIN2018
Archived
4
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

table 2 activity

Browse Source
This commit is contained in:
toni
2018-10-21 00:36:34 +02:00
parent b444319a49
commit 9f77c69ad0
2 changed files with 54 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

65
tex_review/chapters/experiments.tex
View File

@@ -408,34 +408,65 @@ Ironically, this is again some type of sample impoverishment, caused by the afor
\subsection{Activity Recognition} \subsection{Activity Recognition}
\label{sec:eval:act} \label{sec:eval:act}
\add{In order to evaluate the activity recognition, a test person had to press a button according to their current state of motion, namely standing, walking, stairs up, stairs down, elevator up and elevator down (cf. fig. \ref{fig:simple}). \add{In order to evaluate the activity recognition, a test person had to press a button according to their current state of motion, namely standing, walking, walking up, walking down, elevator up and elevator down (cf. fig. \ref{fig:simple}).
As the building does not have an elevator, this state is ignored in the following. As the building does not have an elevator, this state is ignored in the following.
Whether a state needs to be changed was indicated by small symbols on the ground truth markers. Whether a state needs to be changed was indicated by small symbols on the ground truth markers.
This experiment is based on the same measurement series as section \ref{sec:exp:loc}.} This experiment is based on the same measurement series as section \ref{sec:exp:loc}.}
%
\add{As the activity recognition uses moving windows, the detection suffers from a certain lag, depending on their size. \add{As the activity recognition uses moving averages, the detection suffers from a certain lag, depending on their size.
Thus, comparing each activity that is newly calculated with incoming barometer measurements with the ground truth at the current timestamp would result in a rather low detection rate for the respective activities. Thus, comparing each activity that is newly calculated with incoming barometer measurements to the ground truth at the current timestamp, would result in a rather low detection rate for the respective activities.
In addition, only a fraction of a test path consists of the change of an activity, since the testers were walking most of the time. In addition, only a fraction of a test path consists of the change of an activity, since the testers were walking most of the time.
This would bias an overall detection rate. This would bias an overall detection rate.}
It was also investigated, that the standing activity could hardly be recognized, because the test persons constantly moved and turned around to look at the exhibits. %It was also investigated, that the standing activity could hardly be recognized, because the test persons constantly moved and turned around to look at the exhibits.
As a result, a proper evaluation of this activity could not be carried out, so we only evaluate the floor changes.} %As a result, a proper evaluation of this activity could not be carried out, so we only evaluate the floor changes.
%Grafik die das zeigt. %Grafik die das zeigt.
\add{In order to be able to make a statement about the quality, we first determined the average (time) lag within a single conducted walk and then shifted the calculated data accordingly. \begin{table}[t]
The lag is given as the (absolute) difference between the timestamp, the activity changes in ground truth and the first timestamp of an interval, given by the size of $\vec{\omega}_\text{s}$, holding the same activity. \centering
Applying this to the measurement series results in an overall floor detection rate of \SI{}{\percent}, with an average lag over all walks of \SI{}{\second} and a standard deviation of \SI{}{}. \begin{tabular}{cccccc}
The single activities ....} %einzelne werte \toprule
& standing & walking & walking up & walking down & overall \\
\midrule
walk 0 & \SI{65.6}{\percent} & \SI{80.9}{\percent} & - & \SI{84.8}{\percent} & \SI{78.4}{\percent} \\
walk 1 & \SI{49.9}{\percent} & \SI{84.1}{\percent} & - & - & \SI{67.5}{\percent} \\
walk 2 & \SI{57.4}{\percent} & \SI{83.5}{\percent} & \SI{83.5}{\percent} & \SI{82.1}{\percent} & \SI{71.7}{\percent} \\
walk 3 & \SI{45.7}{\percent} & \SI{77.5}{\percent} & \SI{85.1}{\percent} & \SI{77.8}{\percent} & \SI{61.3}{\percent} \\
\midrule
overall & \SI{51.4}{\percent} & \SI{81.5}{\percent} & \SI{84.3}{\percent} & \SI{82.1}{\percent} & \SI{67.9}{\percent} \\
\bottomrule
\end{tabular}
\caption{\add{The resulting detection rates provided by the activity recognition for all conducted walks. As the method suffers from a (time) lag, caused by the used moving average, we shifted the measured activity according to the average lag over all walks (\SI{2.96}{\second}). Some cells of the table are empty, because the respective walk did not require this activity.}}
\label{table:activity}
\end{table}
\add{The main reason to utilize such a method was to detect floor changes. \add{In order to be able to make a statement about the quality, we first determined the average (time) lag of the conducted walks and then shifted the calculated data accordingly.
Independent of the detection rate above, the method is able to detect all floor changes of the conducted walks. This does not allow a perfect, but at least fair examination of the detection rate.
This was quantified by comparing ob in JEDEM ground truth interval min. 75 prozent korrekt erkannte aktivitäten sind. } The lag is given as the (absolute) difference between the timestamp, the activity swaps in ground truth (e.g. from standing to walking), and the first timestamp of an interval, given by the size of $\vec{\omega}_\text{s}$, holding the same activity, given by our recognition method.
This provides in an average lag of \SI{2.96}{\second} and a standard deviation of \SI{1.09}{\second} over all walks.
The resulting detection rates can be seen in table \ref{table:activity}.
They are calculated by dividing the number of correctly detected activities with the number of activities given by the ground truth.
The first thing to notice is the bad recognition rate of standing, especially in comparison to the others.
A major impact on this, is the fact, that we encourage the testers to behave as natural as possible, i.e. like a normal visitor of the museum.
As a result, they often turned around or a took a few small steps within the standing sequences, to look at the exhibits.
This behavior is not mapped by the ground truth.
In addition, using only acceleration for detecting might be a bad choice in the first place, as moving the phone, e.g. by putting it in the trouser pocket, will exceed the threshold.
At the end, this leads to the general question, on how to define standing.
Is it a complete standstill or should it allow for a certain degree of freedom?
The answer is always the same, it depends.
As for this museum scenario, the results for detecting the standing activity are not satisfying and a more advanced approach should be considered.}
\add{The detection rates of up and down sind nicht ganz so schlecht, as the main reason ...
all these changes could be detect with}
\add{In average, there are \SI{xx}{\percent} false detected activity changes per tested walk.
This might seem a lot, however they only had an average duration of \SI{}{\second} ($\approx$ a single barometer update).}
%Ende... Finally, it is not possible to derive the smartphone or the user from the recognition rates, which recommends a general use of the approach.
%das brauchen wir glaub nimmer. das oben ist schon nicht verkehrt.
%\add{The main reason to utilize such a method was to detect floor changes.
%Regardless of the detection rate above, the method is able to detect all floor changes of the conducted walks.
%This was quantified by comparing ob in JEDEM ground truth interval min. 75 prozent korrekt erkannte aktivitäten sind. }
%%estimation %%estimation

6
tex_review/review/01_review.txt
View File

@@ -1,5 +1,11 @@
Reviewer #1: Reviewer #1:
-> A short overview over all changes. This text is send to every reviewer. The individual answers follow directly after this text.
The paper presents a smartphone-based localization system using a particle filter to incorporate different probabilistic models. The comments and suggestions as follows: The paper presents a smartphone-based localization system using a particle filter to incorporate different probabilistic models. The comments and suggestions as follows:
1. The authors mention that "a setup-time of under 120 min for the complete building" in abstract. But I don't find any context about the setup-time in the whole paper. How does the "under 120 min" calculate? How long does the navigation mesh for the whole buliding take? How long does the 42 WiFi beacon installation take? How does the measuremnet of the reference points take? etc. The authors should give the details. 1. The authors mention that "a setup-time of under 120 min for the complete building" in abstract. But I don't find any context about the setup-time in the whole paper. How does the "under 120 min" calculate? How long does the navigation mesh for the whole buliding take? How long does the 42 WiFi beacon installation take? How does the measuremnet of the reference points take? etc. The authors should give the details.