Deep Learning For Image Analysis And Mining Of Biosensing Data

Deep learning algorithms have made extraordinary progress in the classification and recognition of two-dimensional images,such as image classification and recognition,face recognition and other fields.The precise processing and analysis of biosensing data is of great significance to life science research.Based on the above research background,the biosensing data processing is related to deep learning.This paper proposes to convert one-dimensional biosensing data into two-dimensional images,and realize training and learning based on the convolutional neural network model of deep learning.The processed images are classified and recognized,so that observations can be made more intuitively,deeper information can be explored,and the relevance of various information is better.Based on the excellent performance of the deep learning algorithm in image classification and recognition,this paper chooses the convolutional neural network model,chooses the VGG16 model to implement migration learning,extracts the multi-dimensional features and then performs the dimensionality reduction through the principal component analysis algorithm,and obtains the reduced visualization result.Finally,the designed classifier is added to obtain the specific classification of the biosensing data experiment.First of all,this paper uses the data collected during the preparation of the optical fiber biosensor as the research object to conduct experiments,and image processing on it.After feature extraction and dimensionality reduction and visualization,it can be seen that the preparation of optical fiber biosensors can be obvious after the learning and training of deep learning algorithms.Finally,the fully connected layer and softmax layer of the VGG16 model are improved,and the accuracy can reach87.5%.The ECG signal and the speech signal not only have their own characteristic information in the time domain,they can also be analyzed in the frequency domain after frequency domain transformation.This article selects ECG signals and speech signals for image processing,combined with deep learning for classification and recognition.First,the processing of the ECG signal generates two-dimensional images from the two directions of abnormal interval and abnormal amplitude,and then classifies the images with normal ECG signals.After extracting the features through the VGG16 model and reducing the dimensionality of the PCA algorithm,it can be seen that the two types of images are clearly distinguished in the result image,and the accuracy rates can reach96.15% and 92.98%.Then,the speech signal undergoes a short-time Fourier transform to obtain the corresponding spectrogram,and the speech signals of the two persons are classified,and the accuracy of distinguishing the two persons can be obtained to reach99.62%.Based on the above,it is believed that deep learning algorithms can be effectively applied to image classification and recognition of biosensor data.This article transfers the image knowledge of the pre-trained convolutional neural network to the biosensing data after image processing,and then classifies and recognizes the image,so as to realize the classification and recognition of the original data.Compared with the traditional method of directly inputting the ECG signal To learn from the neural network model,the two-dimensional image method is more conducive to observation,which can effectively distinguish different situations of biosensing data.The analysis of specific influencing factors can also process two-dimensional images into three-dimensional images for image classification and recognition,or can classify biosensing data in more detail for partial spectrum amplification.