Speech Emotion Recognition (SER) plays a pivotal role in enhancing human–machine interactions by discerning emotions conveyed through speech signals. This paper investigates the performance of the TIMNET Emotional Modeling Approach in detecting various emotions, including anger, calmness, disgust, fear, happiness, neutrality, sadness, and surprise. TIM-Net employs temporal-aware blocks to capture temporal emotional cues, amalgamates data from both past and future for contextual richness, and amalgamates features from multiple time scales to accommodate emotional nuances. The evaluation utilizes the SAVEE database and extends to assessing the model's efficacy in noisy environments. Various metrics such as Precision, Recall, F1-Score, and Accuracy are employed for comprehensive evaluation. From results, it is noted that the model exhibits robustness in both clean and noisy environments, maintaining accuracy rates above 68% across varying noise levels.