In real-world scenarios, speech signals are often corrupted by various types of noise, which can significantly degrade the intelligibility and quality of the speech. Noise in such environments is highly non-stationary, changing rapidly and unpredictably. A new algorithm has been created to estimate noise in speech signals affected by rapidly changing and unpredictable non-stationary noise. This algorithm aims to enhance speech quality and intelligibility in challenging acoustic settings. It determines signal presence by comparing the power spectrum of noisy speech to a continually updated local minimum, calculated from past noisy speech power spectra with a look-ahead factor. Its standout feature is its rapid adaptability to swiftly respond to new noise patterns, crucial for maintaining speech clarity and quality in dynamic noise environments. Formal experiments validated its effectiveness compared to other noise-estimation methods when integrated into speech enhancement systems, using metrics like mean squared error (MSE) and signal to noise ratio (SNR). This algorithm significantly contributes to noise estimation and voice enhancement, offering effectiveness in handling highly non-stationary noise and being adaptable and versatile for various applications. This algorithm stands as a substantial advancement in the domains of noise estimation and voice enhancement overall. Its utility extends across diverse applications owing to its proficiency in handling highly non-stationary noise conditions, coupled with its adaptability and versatility.