Deep neural networks (DNNs) excel in pattern recognition but are computationally intensive, limiting their use in low-power applications like IoT sensors and wearable healthcare. IoT devices face strict energy and resource constraints, often processing one-dimensional time series data. This project explores DNN hardware accelerators for energy-efficient edge implementation, focusing on sparse DNN models for IoT sensors. The accelerator, based on a RISC-V CPU, will be scalable, programmable, and adaptable to various DNN models. It will be demonstrated on an FPGA (or ASIC) in a wearable biomedical application, such as heartbeat classification.