The explosive expansion of heterogeneous services in Fifth Generation (5G) wireless networks necessitates intelligent resource scheduling techniques that can optimize numerous conflicting performance objectives simultaneously. Traditional scheduling techniques fail to explore the full capabilities of intelligent resource allocation due to static parameter configurations and single objective optimization, leading to poor performance in dynamic network environments. A Deep Q-Network (DQN) based framework was proposed for multi-service resource scheduling in 5G networks. The DQN framework addresses resource allocation problems for Voice over Internet Protocol (VoIP), Video, and File Transfer Protocol (FTP) services. The proposed approach implemented a four-dimensional state space consisting of normalized system throughput, head-of-line packet delay, packet loss rate, and an index for fairness in packet scheduling, with a twelve-action space for dynamically adapting scheduling parameters. The reward function implemented multi-objective optimization through the simultaneous maximization of throughput and fairness, while minimizing delay and packet loss. Extensive simulation results showed the advantages of the DQN-based approach over the traditional Proportional Fair, Modified Largest Weighted Delay First, and Exponential Proportional Fair algorithms. Performance evaluation with 10 UEs and 25 resource blocks over 150 Transmission Time Intervals showed considerable gains: 22.7% gain in throughput, 31.5% total delay reduction and 10.5% fairness improvement. The framework obtained an overall reward score of 0.9471, indicating 109.7% superiority to the best performing traditional algorithm. The findings supported the usefulness of deep reinforcement learning in solving multi-service resource allocation problems for 5G networks.