Scribd Logo
Upload

Welcome to Scribd!

  • 65 views

    Human Activity Recognition Using Convolutional Neural Network

    Uploaded by

    Kaniz Fatema
    This document discusses using a convolutional neural network (CNN) to identify human activities from accelerometer data collected by smartphones. It proposes directly using raw 3D accelerometer data as input to the CNN without preprocessing. The document outlines related work on deep learning and CNNs for activity recognition. It compares the proposed approach to existing methods and discusses the CNN network architecture. The goal is to recognize activities like walking, jogging, sitting, standing, going upstairs and downstairs.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    Human Activity Recognition Using Convolutional Neural Network

    Uploaded by

    Kaniz Fatema
    65 views19 pages
    This document discusses using a convolutional neural network (CNN) to identify human activities from accelerometer data collected by smartphones. It proposes directly using raw 3D accelerometer data as input to the CNN without preprocessing. The document outlines related work on deep learning and CNNs for activity recognition. It compares the proposed approach to existing methods and discusses the CNN network architecture. The goal is to recognize activities like walking, jogging, sitting, standing, going upstairs and downstairs.

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses using a convolutional neural network (CNN) to identify human activities from accelerometer data collected by smartphones. It proposes directly using raw 3D accelerometer data as input to the CNN without preprocessing. The document outlines related work on deep learning and CNNs for activity recognition. It compares the proposed approach to existing methods and discusses the CNN network architecture. The goal is to recognize activities like walking, jogging, sitting, standing, going upstairs and downstairs.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    65 views19 pages

    Human Activity Recognition Using Convolutional Neural Network

    Uploaded by

    Kaniz Fatema
    This document discusses using a convolutional neural network (CNN) to identify human activities from accelerometer data collected by smartphones. It proposes directly using raw 3D accelerometer data as input to the CNN without preprocessing. The document outlines related work on deep learning and CNNs for activity recognition. It compares the proposed approach to existing methods and discusses the CNN network architecture. The goal is to recognize activities like walking, jogging, sitting, standing, going upstairs and downstairs.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    of 19

    Human Activity Recognition using Convolutional Neural Network

    Supervisor Name :
    Sadia Islam Kaniz Fatema
    Lecturer , Dept of CSE Id : 153402325
    City University
    Abstract
     We construct a Convolutional Neural Network
    (CNN) to identify human activities using the data
    collected from the three-axis accelerometer
    integrated in users’ smartphones. The daily human
    activities that are chosen to be recognized include
    walking, jogging, sitting, standing, upstairs and
    downstairs. The three-dimensional (3D) raw
    accelerometer data is directly used as the input for
    training the CNN without any complex
    pretreatment.
      
    Content
      

     1. Introduction
      
     2. Related Work
      
     3. Comparison With Our Work
      
     4. Work Procedure.
      
     5. Network Architecture
      
     6. Background and Motivation
      
     7. Purpose/ why do we choose this project
      
     8. Scope and Objective
      
     9. Features
      
     10. Resources: Software & Hardware used for development
      
     11. Gantt chart
      
     12. Limitations
      
     13. Benefits
      
     14. References
      
      
    Introduction:
     The widespread usage of portable and wearable smart devices such as
    smartphones and smartwatches has enabled the easy gathering of human
    activity data using various device-embedded sensors.Human activity
    recognition was performed by attaching multiple sensors to the user’s body
    which was cumbersome . However, with the proliferation of affordable smart
    devices, the gathering of user’s activity information have become easy through
    the device-embedded sensors. Various methods for detecting human activity
    using smartphone accelerometer sensors have been proposed.
      
     One of the popular methods for recognizing human activity is through the
    Google Activity Recognition API . Google Activity API can detect user’s
    activities.
      
     In order to address the variability of raw human activity data in human activity
    recognition, we present a more robust one-dimensional (1D) Convolutional
    Neural Network (CNN)- based method that utilizes vector magnitude
    accelerometer

     
    Related Work:

     Deep Learning based Human Activity Recognition.


     LSTM and CNN based Human Activity Recognition
    Comparison With Our Work :

    In our project we will use CNN to detect human activity .


    This CNN will consists multiple complex layer which will
    be able to detect complex pattern.

    To extract characters & digits, we will try both CNN and


    template matching method which both existing paper
    proposed before. Here we will add more layers in CNN to
    get new result. We will also apply traditional ML algorithm
    KNN for extraction. By comparing both methods in terms of
    accuracy we will choose the best one .
    Work Procedure:

     Train the Model: Data collection, data pre-processing, feed these


    data to the model
     Human Activity Recognition: Use that trained convolutional
    neural network model.
     Human Activity Recognition: Use various techniques:
     1. Multi-layer perceptron model
     2. Template matching
     3. KNN algorithm for classification
      
     4. Google vision API
     By measuring the accuracy we will choose the best technique for
    Human Activity Recognition .
    Network Architecture:
    Background and Motivation:

     Investigators easily catch criminals for detecting user activities


    such as running, walking and staying.
    Purpose/ why do we choose this project:

     
     To make a business profit .
     For reducing risk of many disease such as cardiovascular, obesity
    and diabetes.
     Globally perspective.
    Scope and Objective:

     Model based on user activities which help detecting criminal


    offenses based on their activities
     It can also apply business for targeted investigators.
     Model will be implemented in a daily activities which build for
    this project.
     For healthcare

     
    Features:

     Model will help for business for the targeted users.


     This model will make good environment for the user .
     Human- Human interaction and interpersonal relations.
     Identify a person , their personality and psychological state.
     
    Resources:
     Software & Hardware used for development:
    Hardware:
     Intel Core i5 2.9 GHz
     2. 8GB Ram
     3.Windows 10 64 bits
    Software:
     Anaconda
     Jupiter Notebook
     
    Limitations:

     Lack of data for training model


     Lack of data security
     It constrained the number, locations and nature of used sensors.
     Deployment, maintenance and costs of daily activities
    unimpeded.
    Benefits:
     Can make a good business profit.
     For detecting criminal offenses
     Great impact on investigators and cop
    Conclusions:
     We presented a CNN-based human activity recognition method
    that uses triaxial accelerometer data gathered from the user’s
    smartphone. Our method outperformed the baseline random
    forest method in ternary human activity classification and
    exhibited the best classification accuracy when longer length
    accelerometer data was used for learning the neural network. We
    found that the dimension of the input vector affects the activity
    recognition performance and that figuring out a way to
    disambiguate the ‘walk’ signal in particular will likely lead to the
    improvement in the activity recognition performance.
    References
     [1] Human Activity Recognition From Accelerometer Data
    Using Convolutional Neural Network,2017Author (Song-Mi
    Lee, Sang Min Yoon, Heeryon Cho).
     [2] Human Activity Recognition Based On Convolutional
    Neural Network,2018, Author (Wenchao Xu, Yuxin Pang,
    Yanqin Yang).
     [3]. Comparing CNN and Human Crafted Features for
    Human Activity Recognition,2017, Author(Federico
    Cruciani).

    You might also like