An intelligent system is proposed to locate and classify audio source signals in large spaces. The system is composed of a sparsely distributed microphone array and an artificial intelligence (AI) system. A sparse array aimed for acoustic source localization and signal extraction is configured. The localization method is based on time difference of arrival (TDOA). This method begins with estimation of the TDOAs among at least 4 sensors, with the aid of a subspace-based time delay estimation algorithm. Next, a constrained least squares (CLS) algorithm is applied to locate the source in accord with the estimated TDOAs. Once the source is located, the source signal is extracted by using the minimum variance distortionless response (MVDR) beamformer and a postfilter. The extracted audio signals are further classified in light of machine learning. Convolutional Long Short-Term Memory (ConvLSTM) plays a central role in the AI-based classifier. Mel-Frequency Spectral Coefficients (MFSC) serves as the input layer in the ConvLSTM. The performance of the proposed system is quantified by using localization error, audio quality, and F 1 scores. Simulations are undertaken to validate the proposed TDOA-based localization and separation technique.