The identification of peaks in time series data, known as peak detection (PD), holds great significance as it pinpoints notable fluctuations within the dataset. These peaks serve as crucial indicators of transitions or anomalies in the time series. This technique finds utility across various applications including spectroscopy, biomedical image processing, and noise signal differentiation, among other domains. The challenge lies in autonomously estimating parameters for a generalized online peak detection algorithm due to the dynamic nature of real-time time series data. Moreover, automating parameter estimation is a major obstacle, given that different domains require adjustments on varying scales. Addressing this challenge, we propose A-TSPD (autonomous-two stage peak detection), a novel algorithm building upon the TSPD (two stage peak detection) algorithm. A-TSPD aims to achieve autonomous parameter estimation, overcoming the need for manual configuration. Experimental validation on seven real-world datasets across diverse domains demonstrates the efficacy of A-TSPD, showcasing significant performance improvements compared to other established techniques.

时间序列数据中峰值的识别（称为峰值检测 (PD)）具有重要意义，因为它可以查明数据集中的显着波动。这些峰值是时间序列中转变或异常的关键指标。该技术可用于各种应用，包括光谱学、生物医学图像处理和噪声信号区分等领域。由于实时时间序列数据的动态特性，挑战在于自动估计广义在线峰值检测算法的参数。此外，考虑到不同领域需要不同规模的调整，自动化参数估计是一个主要障碍。为了解决这一挑战，我们提出了 A-TSPD（自主两级峰值检测），这是一种基于 TSPD（两级峰值检测）算法构建的新颖算法。 A-TSPD旨在实现自主参数估计，克服手动配置的需要。对跨不同领域的七个真实世界数据集的实验验证证明了 A-TSPD 的功效，与其他已建立的技术相比，显示了显着的性能改进。