Background: The physiology of respiration is modulated by autonomic efferent neurons and circulating hormones. Objective: We aim to compare the respiratory patterns in normal individuals, regular meditators and individuals with anxiety spectrum disorders, in an effort to automate the anxiety detection (AnD). Methods: Prospective cross sectional study which included 52 subjects, 20 normal population, 16 regular meditators and 16 subjects with anxiety disorders, was held at the Department of Neurology, Amrita Institute of Medical Sciences (AIMS), Kochi, Kerala, India (IEC: AIMS/2013/18). A 24 hours ambulatory monitoring of each subject was done during the sleep-wake cycle using the respiratory inductance plethysmography (RIP). We evaluated Respiratory Rate Variability (RRV), Thoraco-abdominal ratio (TAR%) and low frequency-high frequency ratio (LF/HF) for its effectiveness in AnD using p-values. Results: It was observed that RRV was lowest, TAR% and LF/HF ratios were highest in anxiety group compared to the meditation group (p=__ampersandsignlt;0.001). Further, we developed a AnD system with statistical features derived from the respiratory signal, and LF/ HF ratio as its input, with a support vector machine (SVM) backend classifier. With the use of efficient signal processing algorithms to remove the effect of patient-specific variations in the feature vectors derived from the respiratory signal, we were able to obtain a performance accuracy of 92.30% absolute. Conclusion: The study highlights that we can automate AnD, and thus minimize the effect of subjective factors in assessing AnD and consider the LF/HF ratio as a new surrogate marker for autonomic imbalance.