ICENS2019, 12 - 16 June 2019
In this study, methods related to amplitude thresholding that are one of the processing steps of Brain Machine Interface (BMI) applications are compiled. The BMI measures the change in the activity of brain cells to any stimuli. Then, it can also extract behavioral data corresponding to this measurement without behavioral data. The running of the BMI can be summarized as follows: Firstly, the neural activity record is filtered interval a band of included of action potential (AP). The AP candidates contained in the filtered record are determined by amplitude thresholding method. Suprathreshold waveforms are classified and the AP sequences that are generated by individual neurons are determined. AP sequences model as a function of behavioral variables, so it is found out what kind of information is encoded in the activity record. By using the activity models in the decoders, the behavioral data are extracted from the activity record. These processes are speed-limiting due to the pattern recognition and classifying that must be performed on hundreds of electrodes. Whether or not password solution can be performed without these processes is trend research topic at last few years. The optimization of the amplitude threshold is one of the most important subheadings of this topic. In literature, the amplitude threshold is calculated as three to five times the standard deviation of the signal. Another method commonly used to calculate amplitude thresholds is to set a threshold between three and five times the RMS voltage of the signal. So, the amplitude threshold should be calculated as user-independent, faster and data-driven. To overcome these disadvantages, an approach based on the modeling of the probability distribution of the signal, called the Truncation Thresholds, has been proposed and proved to be superior to other methods in the literature. Thus, contribution to literature has been provided for BMI.