Introduction Human skin emits a variety of volatile organic compounds that are perceived as unique body odor [1]. It has been believed human body odor contains a large amount of information related to body chemistry of individuals. During recent years, there has been great interest in aging-associated body odors since some compounds have been identified existing in specific age groups [2]. For example, diacetyl has been found is a key contributor for middle-aged male odor [3]. 2-nonenal has been detected only in people over the age of 40 (old person smell) [4]. Organic acids (such as hexanoic acid and its derivatives) are characteristic odor of young ages due to their active activity of eccrine sweat glands [2]. The detection and recognition of aging-associated body odor may suggest many potential applications. However, the assessment of body odor at present mainly depends on human panelist and lack of effective sensing approaches. In this work, a flexible and printable chemiresistor sensor is developed for the detection and recognition of aging-associated odor compounds. The fabrication process of the sensor, the selection method of sensitive materials, and the pattern recognition of an optimized sensor array on three typical odor substances (hexanoic acid, diacetyl and 2-nonenal) are introduced. Materials and Method The chemiresistor sensor was prepared by using conductive composites of carbon black (CB) with various sensing materials. Many materials have been reported that can be used in the fabrication of CB-based chemiresistor sensors including polymers, lipopolymers, low volatility organic molecules, intrinsic conducting polymers, gas chromatography stationary materials, and so on. How to select those materials that are suitable for the sensing of body odor is a problem. In this work we investigated and compared the performance of about fifty materials commercially available. The chemiresistor sensors were prepared by mixing CB and the sensitive materials in tetrahydrofuran (or chloroform) with appropriate critical concentrations. Interdigitated array (IDA) electrode was prepared by a commercial inkjet printer using Ag conductive ink and polyethylene terephthalate (PET) film. The CB suspensions of different sensing materials were dropcast automatically by a digital dispenser on the IDA electrode to fabricate the array sensors. The performance of the sensitive materials was evaluated by their response on five McReynolds test probes (benzene, 1-butanol, 2-pentanone, nitropropane and pyridine). The selectivity of sensor channels were analysis by principle component analysis (PCA) as well as multivariate analysis of covariance (MANCOVA). A 6-channel sensor array optimized according to F-values of Wilk’s Λ statistic was used for pattern recognition of hexanoic acid, diacetyl and 2-nonenal. Vapors with standard concentration (from several ppm to several tens of ppm) were generated by a calibration gas generation system (GASTEC, Japan). The response of chemresistor sensor was recorded by a home-made multiple channel sensing device and control software system. Results and Conclusions Figure 1 shows the illustrated experimental setup used for the generation and sensing of odor samples with standard concentrations. The flexible IDA electrode with 8 channels basically was prepared by a low cost inkjet printing process. The McReynolds probe compounds are generally used in gas chromatography to assess the separation capability of gas chromatography stationary phase. Here we use them as a reference to select the sensitive materials for the fabrication of sensor array. Figure 2a shows the response characteristics of twenty-four materials on the five McReynolds probes (The details of these materials was not disclosed in view of business consideration). Figure 2b shows the ordered F-value of the above 24 materials by analysis of MANCOVA. A large F-value means that the variable has high selectivity [5]. That is to say, the materials holds high discrimination ability in the array. The chemiresistor sensor array can be then optimized by comprehensive consideration both the selectivity (F-value) and sensitivity (response intensity over 5 ‰). For example, the red mark in figure 2b lists an optimized sensor array including 6 channels ( No.13, 34, 32, 20, 23 and 16). Figure 3 shows the PCA result of the optimized sensor array on the McReynolds probes. The five compounds are well separated in the PCA plot which demonstrates that the optimized sensor array may exhibit a broad range of ability to discriminate odorant molecules. This result provides us a guiding principle to design sensor arrays for different sensing targets. The sensor array is then used for sensing of hexanoic acid, diacetyl and 2-nonenal. The PCA plot shown in figure 4 demonstrates that it succeeds in the detection and recognition of aging-associated human body odors. Considering the excellent body odor sensing performance as well as printable and flexible characteristics, it is promising that the chemiresistor sensor array can be further developed as a wearable device and applied for body odor-related biomedical and health application.