Table 1 Features and challenges of conventional movement recognition models in AAL.

Passias et al.³²

Spiking Neural Networks (SNN)

• This method lies in its significant ability for real-time and energy-efficient processing in the AAL environment

• It is unable to reproduce complex neural behaviors

Shah et al.³³

MLSTM

• It is used to learn complex and abstract sequential data representations

• It requires a vast amount of training data

Yadav et al.³⁴

Machine Learning

• It is supportive of making healthcare decisions

• It is hard to interpret and has computational burdens

Jain et al.³⁵

Bi-conventional Recurrent Neural Network

• It obtains high accuracy and controllability

• It is computationally slower than the other network architectures

Pandya et al.³⁶

Bi- LSTM

• It has offered a better performance in multiple conditions

• It becomes slow to train the approach with large data sets

De et al.³⁷

MCP

• It is a low-cost and flexible system

• It takes more processing time and provides inaccurate solutions

Zhang et al.⁴³

DBN

• The result clarifies the efficacy of DBN-TSK-FC and its learning approach

• The architecture of DBN is less flexible and demands hardware to implement DBN

Guerra et al.⁴⁴

LSTM

• The proposed method shows a good balance among the requirements for improved classification sensitivity and accuracy

• It is expensive and requires a large amount of data