Chinese Quince (Cydonia oblonga Miller) Freshness Rapid Determination Method Using Surface Acoustic Wave Resonator Combined with Electronic Nose

Abstract

In this article, Chinese quince (Cydonia oblonga Miller) freshness determination method was investigated using surface acoustic wave resonator, electronic nose, and surface acoustic wave resonator combined with electronic nose. Human sensory evaluation and weight loss index were examined as freshness reference. Results indicated that quince freshness decreased during storage procedure. Surface acoustic wave resonator output frequency and electronic nose measurement data stochastic resonance signal-to-noise ratio Eigen values characterized quince quality under different storage time. Freshness predictive models were developed using surface acoustic wave resonator frequency, electronic nose signal-to-noise ratio spectrum Eigen values, and their hybrid model. Validating experiments results demonstrated that the hybrid predictive model presented higher predicting accuracy (R² = 0.987) than other two models. The proposed method is promising in fruit quality rapid analysis.

Keywords:

• Chinese quince
• Freshness rapid analysis
• Surface acoustic wave resonator
• Electronic noise
• Stochastic resonance

INTRODUCTION

Chinese quince (Cydonia oblonga Miller), is named Lingnan fruit king for its particular nutritional contents (such as sugars, crude fibers, organic amino acids, and minerals), attractive fragrance, and delicious taste.^[1,2] Its unique functions, such as sweet taste, stomach nourishing, and spleen fortifying, have been confirmed.^[3,4] Quince can also be used in drugs. It contributes to weight loss as well as promoting the secretions of breast hormones. However, it is highly perishable after harvest due to environmental factors (such as humidity, temperature, and respiratory metabolism).^[2,5] Quality decline of quinces brings huge economic losses to farmers every year. Previous reports mainly focus on changes in antioxidant properties and storage stability,^[1] quality comparison between different cultivars,^[2] quality preservation after harvest,^[6] and other areas.^[7–14] Little information about quince freshness rapid determination has been reported. A reliable quince freshness rapid determination technique is badly needed.

Non-destructive analysis techniques, such as visible/near infrared (VIS/NIR) spectroscopy, surface acoustic wave (SAW), and some electronic detection devices (electronic noise; EN, and electronic tongue [e-tongue]), have been applied in fruits and vegetables quality analysis.^[15–19] Clark et al. proposed quality predictive models for kiwifruit dry matter, soluble solids content, and flesh color using VIS/NIR spectroscopy combined with canonical discriminant analysis (CDA).^[20] Russo et al. studied red onion ecotypes flavor evaluation by means of EN and discriminant function analysis (DFA).^[21] Campos et al. successfully discriminated grape ripeness in different cultivars using voltammetric e-tongue combined with principal component analysis (PCA). Besides, total acidity and sugar content predictive models were developed using e-tongue combined with partial least squares (PLS) regression.^[22] Hong et al. proposed a new method for adulteration of cherry tomato juices using EN and e-tongue. Meanwhile, pH and soluble solids content predictive models based on fusion approaches showed higher forecasting accuracy than the models using sole usage of EN or e-tongue.^[23]

However, not all these approaches are suitable for practical applications. VIS/NIR spectroscopy is suitable for laboratory analysis.^[24,25] E-tongue is fast, simple, and reliable, but it is limited in sensitivity.^[23,26] The detecting objective of EN technique is the gases emitted by the samples. This technique has diverse applications in chemical analysis,^[27–29] food examination,^[30–32] and clinical medicine.^[33,34] SAW technique developed fast with the development of materials science and electrical engineering in recent years.^[35–37] Previous reports with SAW mainly focus on physical research, odor sensing, and bacteria growth monitoring.^[38–41]

In this article, Chinese quince freshness determination method using surface acoustic wave resonator (SAWR) combined with EN was investigated. Human sensory evaluation (HSE) and weight loss index were examined. SAWR output frequency and EN measurement data were measured. Freshness predictive models were developed using SAWR frequency, EN signal-to-noise ratio (SNR) Eigen values, and their hybrid model. Results indicated that the hybrid predictive model had higher predicting accuracy.

MATERIALS AND METHODS

Quince Samples

Fresh quince samples, in the same size and maturity degree, were bought from Gouzhuang fruit wholesale market (Hangzhou, Zhejiang). All samples were free of any pretreatment or any mechanical damage. After transporting to laboratory, samples were stored at 4℃. Each day, 20 samples were randomly taken for HSE, weight loss, SAWR, and EN measurement. Experiments lasted for 11 days. One hundred new samples were stored at the same storage condition for the validating experiments. All experiments were conducted under room temperature and standard atmospheric pressure.

HSE

HSE has been widely applied in food quality assessment. In this research, sensory evaluation is conducted by six experienced panelists. Voting number is set at k, k∈(1,10). The quality of quince is divided into m levels and the score of a specific level is set at h, j∈(1,m). Quince attributes are divided into n elements, and a specific element is set at u, i∈(1, n). The contributory weight is determined by pairwise comparison of contribution weight of attributes is set at x_i (∑x_i=1). If there is a specific relationship between two objects of h and u, the relation set (matrix) of f is calculated as follows:

(1)

The overall acceptability of quince is calculated by the weight grade method as follows:

(2)

Sensory evaluation scheme for quince samples is displayed in Table 1. Each index is scored from 1 to 5. The higher preference score, the better quality of quince. Scores of three are set as the limit of overall acceptable freshness.

TABLE 1 Sensory evaluation scheme for quince samples

Attributes	Attribute degree
	5	4	3	2	1
Outward integrity	Firm	Integral	General integral	Slight crack	Much crack
Touch	Very hard	Hard	Slight soft	Soft	Very soft
Peel color	Green	Slight yellow	Yellow	Dark yellow	Brown
Shrinkage degree	None	Minor	Modest	large	Total
Glossiness	Very bright	Bright	Slight dull	Dull	Very dull
Rot degree	No rot	Little rot	Some rot	Major rot	Total rot

Weight Loss

Mettler Toledo AL104 electronic balance is utilized to measure quince weight. Weight loss percentage is expressed as the percentage weight decrease respecting to the initial weight.

SAWR System and Measurement

A 433 MHz SAWR is utilized for measurement. ST cut type quartz is used as the piezoelectric base material (see Fig. 1a). Detecting system consists of SAWR and its load circuit, DC power supply (DF1714SB3A, Ningbo CSI Electronics Co., Ltd), and universal digital counter (EE3386, Jiangsu New Union Technology Co., Ltd.; see Fig. 1b). Equivalent model of SAWR in serial with quince is displayed in Fig. 1c, where Co is a static capacitor, Ls is the dynamic inductance, Cs is the dynamic capacitance, Rs is the dynamic resistance of SAWR. When quince is connected to SAWR, Ce is equivalent dynamic capacitor and Re is an equivalent dynamic resistance of sample. The frequency of SAWR loaded with quince sample can be calculated by following equation:

(3)

FIGURE 1 SAWR system: (a) schematic diagram of SAWR system; (b) detecting system; (c) equivalent model of SAWR in serial with quince sample.

In Eq. (3), SAWR’s unloaded frequency , amplification circuit phase parameter , analyte’s conductivity , electrode capacitance , where is a permittivity, and is parasitic capacitance between wires. These parameters keep highly stable, so , , and become decisive factors to oscillation frequency. So, if quinces of different quality are connected to the circuit, the changeable parameters (including , , and) induce the changes in SAWR frequency. SAWR frequency measurement is performed by following procedure: Connect quince with SAWR, measure frequency value by universal digital counter, and measurement data is transferred to a desktop via RS-232 communication interface.

EN System and Measurement

EN system has three main parts: data acquisition, modulating and transmitting unit (U1), sensor array and chamber unit (U2) and power and gas supply unit (U3; see Fig. 2). EN consists of eight metal oxide semiconductors (MOS) gas sensors with different sensitive species. The selectivity toward volatile compound classes of MOS sensors is indicated by the manufacturer: S1 (TGS-825, hydrogen sulfide), S2 (TGS-821, hydrogen), S3 (TGS-826, ammonia), S4 (TGS-822, ethanol, methylbenzene, xylene gas), S5 (TGS-842, hydrocarbon component gas), S6 (TGS-813, methane, and butane), S7 (TGS-2610, propane), and S8 (TGS-2201, nitrogen oxides). Sensor responses are sampled as voltage (V). The MOS sensors rely on changes in conductivity induced by the adsorption of molecules in the gas phase and on subsequent surface reactions. They consist of ceramic substrate coated by metal oxide semiconducting film, and heated by wire resistor. Due to the high temperature (250–500^oC), the volatiles transferred to the surface of the sensors are totally combusted to carbon dioxide and water, leading to a change in the resistance. The high temperature avoids water interference and provides sensors fast response and rapid recovery time. Each sensor is installed in independent chamber to avoid cross-influence of gas flow. Polytetrafluoroethylene (PTFE) material is utilized to fabricate gas sensor chamber.

FIGURE 2 Schematic diagram of EN detecting system.

In EN measurement, quinces were placed into glass air-tight vials, and sealed with sealing membrane. When the power was on, washing pump and valve 2 were started, and sampling pump and valve 1 were close. The air was filtered by active carbon to obtain zero gas. Sensor S1~S8 were washed by zero gas. When the sensor’ response reached the baseline, the washing pump and valve 2 were closed. Then the sampling pump and valve 1 were started. The EN inhaled the headspace gases of samples through the sensor array at 400 mL/min for 40 s. The measurement interval of EN was 0.05 s. When measurement was finished, the sensors were washed by zero gas for 600 s at a flow rate of 1000 mL/min prior to the next sample measurement.

Stochastic Resonance (SR)

SR theory develops fast in recent years.^{[18,19,25,42–44]} This model realizes two functions: noise reduction and signal amplification. It has wide applications in signal processing,^[18,19,25] biochemistry research,^[45] and electrochemical studies.^[46,47] SR model can be defined as following:

(4)

whereis the position of the Brownian particle, is the time, M and C are adjustable parameters, is an input signal with an intrinsic noise , is the external noise, and is the double-well potential with the constants and .

(5)

Equation (4) can be transformed as:

(6)

The minima of are located at , where . A potential barrier separates the minima with the height given by . The barrier top is located at . When three elements of SR interact coherently, the potential barrier can be reduced and the Brownian particle may surmount the energy barrier and enter another potential well. The intensity of signals will increase, which makes it possible that the weak signal can be detected from noise background. Suppose the input signal is , where is signal intensity, is signal frequency. is external noise intensity. SNR is the common quantifier for SR and it can be approximately described as:

(7)

Noise intensity is a parameter in SR model. This model is used for e-nose data analysis. denotes an input matrix. It has a sinusoid signal , EN response data , and intrinsic noise . SNR between the output and input is calculated. This model has been successfully used in food analytical applications.^[32,33] The schematic diagram of SR data analysis is displayed in Fig. 1b.

RESULTS AND DISCUSSION

Sensory Evaluation

Preference scores of quince during storage are presented in Fig. 3a. The initial score is 5, suggesting that the overall sensory acceptability of quinces is the best. With the increase of storage time, preference score decreases fast. It reaches 2.88 ± 0.11 on day 7, which is less than 3 and regarded as unacceptable. After that, it continuously decreases with the increase of storage time. Results indicate that quinces maintain acceptable freshness within 6 days. The sensory changes may be explained by the effect of enzymatic activities and storage environment (e.g., oxygen). Mirror results are reported by Legua et al.^[2] and Ali et al.^[6]

FIGURE 3 Quality examining results: (a) human sensory evaluation; (b) weight loss.

Weight Loss

Figure 3b displays the quince weight loss examination results during storage. With the increase of storage time, weight loss percentage of samples gradually increases, and it reached values of 5% on day 11. Linear fitting between weight loss index and storage time is conducted (). Results indicate that the weight of quinces continues to decline. The reason lies in the moisture loss and inner biochemical content changes of the samples.

Freshness Predictive Model Using SAWR

SAWR frequency measurement results are displayed in Fig. 4. The initial frequency is about 200 MHz. Then the frequency increases to 450 MHz on day 10. According to previous reports, quince physical/chemical indexes change fast after harvest, such as moisture loss, antioxidant capacity, and sugar/acid.^[2,6] The changes alter the internal dielectric characteristics of samples. Due to the different dielectric characteristics of quinces under different freshness, SAWR frequency is significantly influenced when quince is in serial with SAWR circuit. According to Eq. (3), SAWR frequency increases due to a significant increase in Re and decrease in conductivity (). Although the dynamic capacitor parameter (Ce) also changes during storage, it has weak impact on SAWR frequency than Re. So change of Ce can be neglected. To evaluate relationship between SAWR frequency and quince quality, linear fitting regression between SAWR frequency and storage time are conducted. Results are displayed in Fig. 4 and Eq. (8).

(8)

FIGURE 4 Freshness predictive model using SAWR frequency output values.

After one-step transform, Eq. (9) is used as quince freshness predictive model. The input is the frequency, and the output is the storage time.

(9)

Freshness Predictive Model Using EN

Before experiments, EN responses to blank sample are measured during the experiments to evaluate EN stability. Zero gas, obtained by filtering air using active carbon, is used as the blank sample. Results indicate that the responses of sensor array to blank sample are relatively weak (close to zero) and stable. So, sensors’ time drift has no impact on EN stability.

Figure 5a displays EN original responses to quinces on day 0. All sensors’ initiative responses are close to zero. Then the responses increase and reach their stable state. The sensors present different stable values. Sensor S4 has the maximal value (about 0.02 V). The volatile gases emitted by quinces are inhaled into EN gas chamber and sensed by the functional materials equipped within the sensors. The specific absorption from these materials for specific gas species induces materials’ changes in their electrical characteristics. And the responses increase with the increase of gas concentrations. EN sensor array forms different pattern signals toward samples under different storage time, and the stable responses are selected for the subsequent analysis.

FIGURE 5 EN analysis results and freshness predictive model based on EN system: (a) EN sensor array original responses to quince samples; (b) PCA score plot; (c) SR SNR spectrum; (d) freshness predictive model.

Figure 5b shows the projections of PCA results on a two-dimensional plane by the first two principal components. The first two principal components, PC1 and PC2, totally capture 84.94% of data variance. PC1 and PC2 first increases and then declines with the increase of storage time. The increase of storage time and day 6 was a transition point. Quinces under the same storage time gather together. Samples with different freshness can be discriminated by PCA. However, this method just provides a qualitative analysis method, not suitable for quantitative analysis. Changes in PCA score can be explained by the aroma variance from different quince samples.

Quince SR SNR spectrum as function of noise intensity is presented in Fig. 5c. Noise intensity ranges from 0 to 400. SNR values of all curves first increase with the increase of noise intensity, and then derived vales and peaks appear before the formation of feature peaks. Feature peak located noise intensity is 80. Afterward, SNR values gradually decrease with the increase of noise intensity. Maximal SNR () values of quince samples range from –76.42 to –65.20 dB, and increase with the increase of storage time. values discriminate all quinces under different storage time. Quince freshness predictive model is developed by linear fitting regression between values and storage time (see Fig. 5d and Eq. 10).

(10)

After one-step transform, Eq. (11) is used as the quince freshness predictive model. The input is the EN values, and the output is quinces’ storage time.

(11)

Freshness Predictive Model Using EN Combined with SAWR

SAWR characterizes quince internal quality and EN presents quince external aroma information. Whether a freshness predictive model using EN combined with SAWR presents better predicting accuracy than former two models, Eq. (12) displays the combination model. Two confidence coefficients ( and ) are used in the model. By inputting , , and values into Eq. (12), the values of and are calculated as 0.4552 and 0.5623, respectively. Then the mixed freshness predictive model using fusion data is successfully developed after inputting the two values into Eq. (12), as displayed in Eq. (13). The input is and values, and the output is quinces’ storage time.

(12)

(13)

Validation Experimental Results

To evaluate the accuracy of developed freshness predictive models, validating experiments are conducted. Another 100 quinces are stored at 4°C for 11 days, and 50 samples are randomly taken out for the SAWR and EN measurements. The validation experimental results are displayed in Fig. 6. The regression coefficients () of the three developed models are 0.918, 0.950, and 0.987, respectively. Results demonstrate that the predictive model using EN combined with SAWR presents better forecasting accuracy.

FIGURE 6 Validation experimental results for the predictive model based on (a) SAWR system; (b) EN system; and (c) SAWR combined with EN, respectively.

CONCLUSIONS

In this article, quince freshness predicting method by using EN and SAWR was investigated. HSE and weight loss were examined to monitor the freshness change during storage. Results suggested that sample’s overall acceptability continuously decreased. Responses of SAWR and EN to quinces were synchronously measured. SAWR frequency and SR SNR spectrum feature values quantitatively discriminated quince freshness. Three freshness predicting models were developed using SAWR, EN, and EN combined with SAWR, respectively. Validation results demonstrated that the model using EN combined with SAWR exhibited better forecasting accuracy than other two models. The proposed method successfully predicted quince freshness with good accuracy. And it is promising in fruit quality rapid and non-destructive evaluation. We are carrying out a long-term plan to apply this technique to other fruit and vegetables.

FUNDING

This work is financially supported by National Natural Science Foundation of China (Grant No. 81000645, 61272313), China Postdoc Science Foundation (Grant No. 2014M551749), Natural Science Foundation of Zhejiang Province (Grant No. LQ13F030012), Scientific Research Development Project of Zhejiang A&F University (2015FR020, 2014FR071), and Young Academic Leaders Foundation in Zhejiang Province (pd2013243).

Additional information

Funding

This work is financially supported by National Natural Science Foundation of China (Grant No. 81000645, 61272313), China Postdoc Science Foundation (Grant No. 2014M551749), Natural Science Foundation of Zhejiang Province (Grant No. LQ13F030012), Scientific Research Development Project of Zhejiang A&F University (2015FR020, 2014FR071), and Young Academic Leaders Foundation in Zhejiang Province (pd2013243).