Stacked Progressive Auto-Encoders for Clothing-Invariant Gait Recognition

Yeoh, TzeWei; Aguirre, Hernán; Tanaka, Kiyoshi

doi:10.1007/978-3-319-64698-5_14

Cited by 3 publications

(3 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare several state‐of‐the‐art deep learning methods for gait recognition, including Auto‐Encoder, 4 SPAEGait, 5 GEiNet, 10 DeepCNNs, 11 I/O‐ACNN, 12 Multilayer‐CNNs 9 with our proposed UDAE method. All the input features of different methods are GEI and Auto‐Encoder, SPAEGait+PCA, UDAE are generative models while GEINet, I/O‐ACNN and Multilayer CNNs are discriminative models.…”

Section: Methodsmentioning

confidence: 99%

“…There has a lot of researches based on Convolutional Neural Network (CNN), Generative Adversarial Network (GAN), Multi‐layer Auto‐Encoder, 4 and so on. In 2017, Yeoh et al 5 proposed a Stacked Progressive Auto‐Encoders model (SPAEGait) to directly recognize different subjects under clothing variations by transforming the Gait Energy Images (GEI) with complicated clothing into normal clothing types. But the feature extracted in generative model is often used to transform features in different conditions into the same condition, so it would not be class discriminative.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A unified auto‐encoder method for gait recognition under different sensor locations

Chang

Wang

Lin

et al. 2022

Internet Technology Letters

View full text Add to dashboard Cite

Sensor‐based gait recognition has recently been one of the most challenging task in personal identity verification. But the variations of sensor location would affect the quality of collected data, which results in the instability of recognition performance. To tackle with this, this paper proposes a unified discriminative Auto‐Encoder (AE) framework to directly extract discriminative features under different sensor locations, and simplifies the AE‐based framework at the same time. Firstly, we change traditional mean square error of AE into spectral angle distance to keep the geometry of data; Secondly, similar identity code is introduce to extract discriminative features; Moreover, a scatterness regularization is added to ensure the dispersion of the same class. The experimental results show the superiority of the proposed method over other state‐of‐the‐art methods. This paper proposes a unified autoencoder framework for gait recognition based on similarity identity code, which not only reduces the impact of different sensor location, but also simplifies the original progressive autoencoder scheme. The framework includes four steps: sensor‐based data collection, preprocessing, unified autoencoder algorithm and SVM recognition.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A unified auto‐encoder method for gait recognition under different sensor locations

Chang

Wang

Lin

et al. 2022

Internet Technology Letters

View full text Add to dashboard Cite

show abstract

“…In such algorithms, the subsequent frames are analyzed to obtain the user’s pose and location of limbs. In [ 32 ], a model of stacked progressive autoencoders (SPAEs) is employed to convert the gait energy images (GEIs) registered for people dressed in untypical, complicated clothes, e.g., down jacket or long coat, to images of people in typical clothing. The model consists of two stacked autoencoders that aim to map the GEIs of persons in unconventional clothing to typical conditions while keeping the typical GEIs unchanged.…”

Section: Related Workmentioning

confidence: 99%

Wrist-to-Tibia/Shoe Inertial Measurement Results Translation Using Neural Networks

Kolakowski,

Djaja-Josko,

Kolakowski

et al. 2024

Sensors

View full text Add to dashboard Cite

Most of the established gait evaluation methods use inertial sensors mounted in the lower limb area (tibias, ankles, shoes). Such sensor placement gives good results in laboratory conditions but is hard to apply in everyday scenarios due to the sensors’ fragility and the user’s comfort. The paper presents an algorithm that enables translation of the inertial signal measurements (acceleration and angular velocity) registered with a wrist-worn sensor to signals, which would be obtained if the sensor was worn on a tibia or a shoe. Four different neural network architectures are considered for that purpose: Dense and CNN autoencoders, a CNN-LSTM hybrid, and a U-Net-based model. The performed experiments have shown that the CNN autoencoder and U-Net can be successfully applied for inertial signal translation purposes. Estimating gait parameters based on the translated signals yielded similar results to those obtained based on shoe-sensor signals.

show abstract

Deep learning pipelines for recognition of gait biometrics with covariates: a comprehensive review

Parashar

Parashar²,

Ding

et al. 2023

Artif Intell Rev

View full text Add to dashboard Cite

Stacked Progressive Auto-Encoders for Clothing-Invariant Gait Recognition

Cited by 3 publications

References 13 publications

A unified auto‐encoder method for gait recognition under different sensor locations

A unified auto‐encoder method for gait recognition under different sensor locations

Wrist-to-Tibia/Shoe Inertial Measurement Results Translation Using Neural Networks

Deep learning pipelines for recognition of gait biometrics with covariates: a comprehensive review

Contact Info

Product

Resources

About