Spike sorting is one of key techniques to understand brain activity. With the development of modern electrophysiology technology, some recent multi-electrode technologies have been able to record the activity of thousands of neuronal spikes simultaneously. In this scenario, however, the recorded activity may be the overlap of multi-neuron spikes, which will degrade the sorting performance of existing cluster-based algorithms. In this paper, we introduce methods for overlapping spike sorting. The introduced methods start from a convolution model, where a sparse vector could be obtained via sparse coding or compressive sensing. Then, we use a maximum a posteriori (MAP) estimate to optimize the sparse vector, which make the overlapped spike sorting completed successfully. The advantage of the introduced method is that, it performs better than traditional methods when the waveforms of the spikes are similar. In experiments, some synthetic and real spike data are used to testify the methods. The experiment results show that the introduced methods’ average sorting detection, defined as the ratio of successfully sorted spikes to the total spikes is nearly 4% higher than traditional methods, under the condition of the experimental data with similar waveforms.

Spike sorting is one of key technique to understand brain activity. With the development of modern electrophysiology technology, some recent multi-electrode technologies have been able to record the activity of thousands of neuronal spikes simultaneously. The spike sorting in this case will increase the computational complexity of conventional sorting algorithms. In this paper, we will focus spike sorting on how to reduce the complexity, and introduce a deep learning algorithm, principal component analysis network (PCANet) to spike sorting. The introduced method starts from a conventional model and establish a Toeplitz matrix. Through the column vectors in the matrix, we trains a PCANet, where some eigenvalue vectors of spikes could be extracted. Finally, support vector machine (SVM) is used to sort spikes. In experiments, we choose two groups of simulated data from public databases availably and compare this introduced method with conventional methods. The results indicate that the introduced method indeed has lower complexity with the same sorting errors as the conventional methods.

Capture effect is very common in wireless communication systems. In a passive radio frequency identification (RFID) system, even though multiple tags backscatter their signals to a reader simultaneously, one of the tags will be successfully identified due to the capture effect. In this letter, we propose an algorithm which will optimize the identification efficiency of a dynamic frame length ALOHA RFID protocol under capture effect. The proposed algorithm estimates the number of tags and the probability of capture effect. From these parameters, an optimal frame length can be obtained. The advantage of the algorithm is that it does not need many steps to search an extreme value for the estimation. It thus reduces the computational complexity. Computer simulation results show that the proposed method's identification efficiency is almost identical to existing algorithms, but at lower computational complexity.

In this paper, we study how to use the number of spike signals in a macaque's motor cortex to estimate the position of its finger movement. First, we analyze the time correlation of a traditional state space model (SSM) and derive a convolutional space model (CSM) to decode the movement position of the macaque finger. Compared with the traditional model, the model can correlate the current moment state with the previous moment. In addition, we have improved the original SSM model using the CSM model. In its observation equation, the number of spike signals at one is expanded to the number of spike signals at multiple previous moments. In this way, the current time state in the improved model can be related to the number of spike signals at a plurality of previous times. In the experiment, a group of public data are used to validate the decoding performance of the model, and a least squares, a batch recursive least squares, and a gradient descent algorithms are used to train the model. The experimental results show that the decoding error of the CSM model and the improved SSM model is smaller than that of the traditional model, and thus, they have a better decoding accuracy. The results show that the CSM model has improved about 11.7% in x-axis decoding error performance than the traditional models.

In a passive ultra-high frequency ( UHF ) radio frequency identification ( RFID ) system, the recovery of collided tag signals on a physical layer can enhance identification efficiency.However,frequency drift is very common in UHF RFID systems, and will have an influence on the recovery on the physical layer. To address the problem of recovery with the frequency drift, this paper adopts a Radial Basis Function ( RBF ) network to separate the collision signals, and decode the signals via FM0 to recovery collided RFID tags. Numerical results show that the method in this paper has better performance of symbol error rate ( SER ) and separation efficiency compared to conventional methods when frequency drift occurs.