Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition


Resource | v1 | created by janarez |
Type Paper
Created 2016-01-18
Identifier ISSN: 1424-8220

Description

Human activity recognition (HAR) tasks have traditionally been solved using engineered features obtained by heuristic processes. Current research suggests that deep convolutional neural networks are suited to automate feature extraction from raw sensor inputs. However, human activities are made of complex sequences of motor movements, and capturing this temporal dynamics is fundamental for successful HAR. Based on the recent success of recurrent neural networks for time series domains, we propose a generic deep framework for activity recognition based on convolutional and LSTM recurrent units, which: (i) is suitable for multimodal wearable sensors; (ii) can perform sensor fusion naturally; (iii) does not require expert knowledge in designing features; (iv) explicitly models the temporal dynamics of feature activations.

Relations

fits under Deep learning

Deep learning (also known as deep structured learning) is part of a broader family of machine learnin...

is about Mobile phone based sensing software

Mobile phone–based sensing software is a class of software for mobile phones that uses the phone's se...

Currently, no authors are attached.

Currently, no resources are attached.


Edit resource New resource

6.5 /10
useless alright awesome
from 2 reviews
Write comment Rate resource Tip: Rating is anonymous unless you also write a comment.
Resource level 6.0 /10
beginner intermediate advanced
Resource clarity 7.0 /10
hardly clear sometimes unclear perfectly clear
Reviewer's background 4.5 /10
none basics intermediate advanced expert
Comments 2
jjones
0 0

8 rating 5 level 8 clarity 4 user's background

Good details, open source
First paper to combine CNN and LSTM for activity recognition
janarez
1 0

5 rating 7 level 6 clarity 5 user's background

Simple deep learning architecture.
Loads of text, but not much of it concentrates on the LSTM part that was innovative in this paper.
This is the first application of CNNs + LSTMs to activity prediction from mobile sensors as such deep learning here is fairly straightforward. A paper for mostly historical purposes.