Scribd
Upload
25 views

IMU Based Localization

The document discusses using inertial measurement units (IMUs) in smartphones to detect indoor location through step counting and stride length estimation. It describes filtering accelerometer data from IMUs to detect steps and determine if the phone is held in the left or right hand. A peak detection algorithm is used to identify steps from gravity vector peaks above a threshold. Stride length is calculated based on the distance between acceleration peaks. The method provides a proof of concept for indoor localization but faces challenges from sensor noise, different carrying positions, and phone movement during activities like running.

Uploaded by

sai k
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
25 views

IMU Based Localization

The document discusses using inertial measurement units (IMUs) in smartphones to detect indoor location through step counting and stride length estimation. It describes filtering accelerometer data from IMUs to detect steps and determine if the phone is held in the left or right hand. A peak detection algorithm is used to identify steps from gravity vector peaks above a threshold. Stride length is calculated based on the distance between acceleration peaks. The method provides a proof of concept for indoor localization but faces challenges from sensor noise, different carrying positions, and phone movement during activities like running.

Uploaded by

sai k
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Indian Institute of Science, Bengaluru

1
IMU Based Localization
IISc
OVERVIEW

 GPS does not work everywhere; indoor is an issue


 Can we use other type of sensors to detect where a

Indian Institute of Science, Bengaluru


person is inside a building?

 Smart phones are sensor rich


 IMUs?
 Develop an algorithm to detect step count and stride
length

2
INERTIAL SENSORS
• Inertial Measurement Units(IMUs) are a
combination of three sensors namely, the

Indian Institute of Science, Bengaluru


accelerometer, gyroscope and
Magnetometer.

• Each sensor outputs data from each of the


three axes(X,Y and Z).
3
Accelerometer data in all three axes
All axes accelerometer data

Indian Institute of Science, Bengaluru


4
Step Detection
Filtered Accelerometer value

Indian Institute of Science, Bengaluru


5
Phone Clasping
•Motion mode Recognition
–For indoor navigation, people can hold phone in any
position like texting, swinging hand, running, carrying

Indian Institute of Science, Bengaluru


phone in bag, while talking over phone etc

•Considered three modes of motion


–Texting mode
–Swinging hand
–Running hand

6
Phone Clasping:
Human Activity Recognition
•Raw accelerometer and gyroscope data is collected from
the phone

Indian Institute of Science, Bengaluru


•This data is converted to a signal image, every signal
sequence has the chance to be adjacent to every other

•The signal is then analysed and processed to get


meaningful insight.

7
STEP DETECTION ALGORITHM

Accelerometer data Filter

Indian Institute of Science, Bengaluru


Filtered IMU
data

Mode Recognition

Right hand/Left Step count/Stride


hand detection length
8
Accelerometer data
 The raw accelerometer data is dynamically collected
from the phone and written into a file
Filters
 Noisy data is smoothened using appropriate filters

Indian Institute of Science, Bengaluru


 High Pass, Low Pass and Band Pass filters can be used
based on the spectrum of signal
Human Activity Recognition
 The filtered data is recorded and the peak detection
algorithm is designed to obtain the peaks
 The individual peaks are detected

 Gravity vector’s peak above a threshold is used to detect


9
the activity
Right hand/ Left hand detection

 When the person holds the phone in right hand:


 swinging towards the direction of motion, the x axis
acceleration peaks are always positive

Indian Institute of Science, Bengaluru


 When a person holds the phone in left hand:
 swinging towards the direction of motion, the x axis peaks
acceleration peaks are always negative

10
STEP DETECTION

 Perform peak detection algorithm for all three axes


 Threshold based analysis is applied to each of the

Indian Institute of Science, Bengaluru


peak and once it is satisfied, the peak is considered as
a step
 Stride length is calculated by taking the distance
between the local maxima and minima of each peak

11
Indian Institute of Science, Bengaluru
12
STRIDE LENGTH
STRIDE LENGTH - CALCULATION

Stride Length = K* (a_max – a_min)

Indian Institute of Science, Bengaluru


 K is a constant derived based on observation.

 a_min is the minimum point when compared Min 1 and


Min 2 of the stride

13
Step count

 The
steps captured is incremented to provide the total
number of steps taken to reach the destination

Indian Institute of Science, Bengaluru


Orientation
 Orientation and Heading information is gathered by
fusing the accelerometer and magnetometer data
 Magnetometer is capable of detecting the fluctuations
in the Earth's field along three perpendicular axes X,
Y and Z
14
STEPS DETECTED IN TEXTING MODE

Indian Institute of Science, Bengaluru


No of steps taken = 20
No of peaks detected = 20
15
CHALLENGES
● The sensor in the phone is noisy to eliminate the noise
suitable filter has to be designed
● Switching from left hand to right hand

Indian Institute of Science, Bengaluru


● False step removal
● While Running or jogging the phone moves randomly and
unevenly in all the directions
– There is noise, the data has to be effectively filtered
● The Orientation is a challenge in Running mode due to
random movement of phone in all the directions
16
● Phone shake must be eliminated

You might also like