Anomaly detection for time series data with low SNR such as the gravitational wave detection problem is challenging because the noise is typically overwhelming. A convolutional neural network (CNN) is a deep learning algorithm efficient especially for image analysis. It has been used for time-series data analysis and proven to be useful for time-series classification as well. The one-dimensional convolution is applied to time-series data and optimal filters are determined through learning. However, one-dimensional CNN for the anomaly detection in time-series data requires a large training set and the accuracy suffers significantly in the case of low SNR. In this talk, we explain that the topological embedding can enhance the efficacy of CNNs for time-series data analysis. Specifically we use topological data analysis (TDA) based on persistent homology. The homological features by embedding of the given time-series data are obtained through TDA and participate in the procedure of CNN. The resulting method is more resilient to noise, more capable of detecting signals with varied signatures and requires less training. This is a powerful improvement as the detection problem can be computationally intense and is concerned with a relatively large class of wave signatures.