Zhejiang University

Advanced New Technologies Co., Ltd.

Yueming Wang

Gang Pan

Yu Qi

Jiliang Li

Zhaohui Wu

Jianmin Zhang

Tao Fang

Dong Xing

Xiaoxiang Zheng

Xiaomeng Yang

Yiwen Wang

Qiaosheng Zhang

Kedi Xu

Cunle Qian

Xinyun Zhu

Yaoyao Hao

Junming Zhu

Disclosed in the present invention is a method for decoding Chinese character writing for an invasive brain-computer interface. In a practical application, a corresponding motor neural signal is divided into two states of a writing stroke and a writing stroke break in view of inconsistency of the writing stroke and the writing stroke break during Chinese character writing, and different filters are trained. A hidden markov model (HMM algorithm) and a Viterbi algorithm are used to judge a task state of the motor neural signal, and the corresponding signal is put into a corresponding decoder. The present invention effectively reduces influence of difference of neural data in different states on the decoder, and improves the performance and robustness of the decoder.

Chinese character writing and decoding method for invasive brain-computer interface

The present invention discloses an adaptive brain-computer interface decoding method based on multi-model dynamic ensemble, where a traditional state-space model is improved, and a set of measurement functions instead of one fixed measurement function are used to dynamically characterize a relationship between observation variables and state variables; and, by using a pool of linear and nonlinear decoders, and in a decoding process of a brain-computer interface system, decoders are automatically switched according to the data, so as to realize adaptive brain signal decoding. Through the above multi-model ensemble strategy, linear and nonlinear decoder capabilities can be integrated, the accuracy and stability of the brain-computer interface system can be improved, and decoding unstability caused by the non-stationary neural signal of the brain-computer interface system can be solved to a certain extent.

Adaptive brain-computer interface decoding method based on multi-model dynamic integration

The present invention discloses a brain machine interface decoding method based on spiking neural network, comprising: (1) constructing a liquid state machine model based on a spiking neural network, the liquid state machine model consists of an input layer, an middle layer and an output layer, wherein, a connection weight from the input layer to the middle layer is W, a loop connection weight inside the middle layer is W, a readout weight from the middle layer to the output layer is W; (2) Inputting a neuron spike train signal, and training each weight with the following strategy: (2-1) Using STDP without supervision to train the connection weight Wfrom the input layer to the middle layer; (2-2) Setting the loop connection weight Winside the middle layer by means of distance model and random connection, and obtaining a middle layer liquid information R(t); (2-3) Using ridge regression with supervision to train the readout weight Wfrom the middle layer to the output layer, and establishing a mapping between the middle layer liquid information R(t) and the output motion information, and finally outputting a predicted motion trajectory. The present invention can quickly train a model in a relatively short time, predict the arm motion trajectory in real time, and achieve an improvement in efficiency and accuracy.

Brain machine interface decoding method based on spiking neural network

The present invention is related to a method for prediction of cortical spiking neural signal, comprising the following steps: 1) pretreatment of spiking neural signal, 2) modeling of posterior cortical spiking neural signal generation probability, 3) model accuracy measurement, 4) model optimization with the help of numerical gradient descent and 5) prediction of posterior cortical spiking neural signal with iterative calculation. The prediction method aims to incorporate natural features of point process of spiking neural signal into the target for optimization of prediction model to improve model's capability in predicting neural spike trains.

Method for prediction of cortical spiking trains

This disclosure relates to model loading. In one aspect, a method includes determining, based on a preset execution script and resource information of multiple execution nodes, loading-tasks corresponding to the execution nodes. Each execution node is deployed on a corresponding cluster node. Loading requests are sent to the execution nodes, thereby causing the execution nodes to start execution processes based on the corresponding loading requests. The execution processes start multiple model service frameworks on each cluster node. Multiple models are loaded onto each of the model service frameworks. Each loading request includes loading-tasks corresponding to the execution node to which the loading request was sent. The execution processes include a respective execution process for each model service framework.

Loading models on nodes having multiple model service frameworks

Growing community of inventors

Hangzhou, China

Yueming Wang