Introduction

Chinese cuisine, as a representative of traditional Chinese culinary culture, is deeply intertwined with culinary practices. It reflects China’s enduring culinary traditions, influenced by diverse regional and ethnic cultures. With a wide array of techniques, ingredients, and regional styles like Sichuan, Cantonese, and Hunan cuisine, Chinese culinary practices are integral to its essence. The key of Chinese cuisine lies whithin techniques such as cutting, stir-frying, deep-frying, steaming, boiling, and braising, aimed at preserving original flavors and achieving a harmonious balance of taste, aroma, and appearance1. As Chinese cuisine gains popularity worldwide, it caters to the global population of 8 billion. Additionally, intelligent Chinese cooking involves automated cooking using intelligent cooking robots and recipe programs to optimize food processing and produce Chinese dishes in a highly autonomous manner2.

Currently, there exists two types of cooking robots: 1. bio-inspired cooking robots based on mechanical arms; 2. specialized cooking robots with optimized mechanical structures tailored to the cooking process. As for the first type, the Moley Robotics cooking robot system3 exemplifies bio-inspired cooking robots, featuring a mechanical arm, electronic ordering system, refrigeration and storage equipment, and a set of kitchen appliances optimized for the robot. This system can prepare over 5,000 dishes and accommodate custom recipes, but its high cost (approximately $335,000) restricts accessibility to ordinary users. Therefore, current research focus on optimizing mechanical structures for Chinese cooking techniques, cost reduction, and developing affordable and efficient cooking products for consumers. Specialized cooking robots include large-scale drum stir-fry pans4, large-scale vertical stirring fryers5,6, and small household automatic stir-fry machines7. The large-scale drum stir-fry pan and large-scale vertical stirring fryer have limitations in achieving fine cooking and large-scale dish processing, resulting in poor dish appearance and shape, making them less appealing to users. Consequently, the small household automatic stir-fry machine is favored by consumers due to its affordability and higher level of automation8.

Due to the complexity of Chinese cooking techniques, the execution of cooking robot programs generates large-scale, multi-dimensional, real-time data, imposing high requirements on the processing capabilities of cooking robots. Therefore, research on the data acquisition devices of cooking machine/robot is essential to provide data support for program execution.

This paper introduces a Chinese cuisine data acquisition device (Chinese Cuisine Intelligent Cooker Recorder-V1.0, CCICR-V1.0) designed to record various pose information, real-time collection, and storage of food status and related information during the cooking process. This includes tracking food entering the pot, ingredients and seasoning added, pot heating and temperature adjustment, pot tilting and rotation, cooked food exiting the pot, pot cleaning, and food insulation9. The recorded food status encompasses food maturity and initial color, while the recorded information quantity includes food type, weight, real-time pot temperature, amount of seasoning added, cooking time, maturity, and stir-fry speed. CCICR-V1.0 comprises a recording system and a control system. The recording system primarily utilizes HX711, MLX90614, MPU6050, OV2640, and LD3320 sensors; while the control platform employs the FreeRTOS system10 based on STM3211 for data collection.

In the context of sensor applications, HX711 is a high-precision 24-bit analog-to-digital converter specifically designed for measureing the outut of voltage and weight sensors. It is commonly used in conjunction with strain gauges or load cells, Chawla et al.12 used HX711 sensors to build a robotic arm for grasping LPG gas cylinders in dangerous scenarios. HX711 has high sensitivity and can meet scene requirements to avoid safety hazards. Lestrai et al.13 used HX711 to construct a sweet potato grasping system. HX711 has high precision and can avoid damage to sweet potatoes. Minera et al.14 used HX711 to build a robotic arm to support visually impaired people’s movements. HX711 has high precision and sensitivity and low cost, making it suitable for building IoT devices and providing services for different groups of people.

MLX90614 is a versatile infrared sensor that can be used for high-precision temperature monitoring. Awwad et al.15 used MLX90614 to track the surface temperature above a leaking pipeline and recorded the corresponding position for each reading, achieving accurate monitoring of leaks based on their thermal anomalies. Rinanto et al.16 improved the recognition algorithm based on MLX90614 hardware for precise measurement of human body surface temperature.

MPU6050 is a digital motion processor that integrates a three-axis accelerometer and a three-axis gyroscope. It can be used to estimate the motion attitude and state of objects and is commonly used in robotics17, agriculture18, healthcare19, and other fields. Li et al.17 used MPU6050 to build a height measurement device for a cutting machine, where MPU6050 was mainly used to measure the roll angle, pitch angle, yaw angle, and height of the drum, achieving high accuracy with low error to meet application requirements. Liang et al.20 built an attitude monitoring system using MPU6050 and optimized the monitoring accuracy through algorithms, further exploring the potential of MPU6050. Sun et al.18 used MPU6050 to build hardware detection devices placed on sows, combined with video multimodal data to achieve precise identification of sow micro-behaviors, improving farm profitability. He et al.19 used MPU6050 to construct a fall detection equipment for continuous and high-precision tracking detection of sudden falls in the elderly.

OV2640 is a 2-megapixel HD image sensor from OmniVision, widely used in smartphones and surveillance cameras. It supports JPEG and RAW outputs and features automatic exposure, white balance, and noise cancellation, ensuring high-quality images in various lighting conditions. Its compact design and low power consumption make it ideal for embedded devices that have space and energy efficiency requirements. Elhattab et al.21 proposes a new model that utilizes ESP32-CAM, OV2640 camera module, a solar panel, and a mobile application to remotely monitor plant growth in agricultural fields, aiming to improve productivity and reduce farmers’ workload. Vardhan et al.22 presents a framework for a UAV-based intrusion detection system using an ornithopter UAV equipped with an ESP32 microprocessor and an OV2640 camera module. The proposed system addresses the limitations of traditional surveillance systems by providing an adaptable and effective solution for detecting intruders in various fields and applications.

LD3320 is a high-performance voice recognition chip widely used in smart homes, IoT devices, and electronic products. It supports various voice recognition functions, including keyword recognition and voice command parsing, enabling offline voice processing and significantly reducing latency. With its low power consumption, the LD3320 is well-suited for devices that require long-term operation. Xiong23 introduces the design of an intelligent trash can that utilizes LD3320 non-specific person speech recognition chip, Arduino UNO R3 MCU, and LX-225 serial bus intelligent steering gear. The system enables voice input of garbage name, intelligent retrieval of garbage type, and intelligent opening and closing of the garbage can cover. Zou et al.24 presents a speed recognition system installed in a smart home. The system allows users to orally operate TVs and lighting fittings using speed recognition technology. By utilizing the LD3320 speech recognition module, infrared T-R module, and STC89C52 single-chip computer, the system achieves intelligent control of home appliances with a high functional accuracy of 93%.

STM32 is widely used due to its strong functionality, low power consumption, and high scalability. Jiang et al.25 used a STM32-based hardware platform running the FreeRTOS system to implement the control terminal for smart homes. The Indian Institute of Technology Madras26 used STM32 with integrated FreeRTOS operating system to build the core control platform for the STUDSAT-2 satellite, enabling ground remote sensing detection27,28. In addition, the Canadian Space Agency29 adopted the STM32+FreeRTOS mode to build the core control for the Canadian CubeSat project initiated in 2018, completing scientific missions related to space weather research. In addition to STM32, embedded MCUs also include AVR, PIC, MSP430, NXP LPC, ESP32, and FPGAs30,31, which are widely used in smart devices, IoT, industrial control, and various other fields. In summary, the hardware and software configurations used in CCICR-V1.0 are mature and reliable solutions that have been validated in other fields.

This paper synthesized the key features and incorporated functions into CCICR-V1.0, presenting the following contributions:

  • Design Objective: CCICR-V1.0 is engineered to address the complexities and data challenges of Chinese cooking, systematically recording relevant data and supporting intelligent cooking machines and robotic systems.

  • Technical Specifications: Featuring STM32 microcontroller and multiple sensors, CCICR-V1.0 enables high-frequency data acquisition and storage, significantly improving accuracy while reducing costs.

  • Commercial Viability: CCICR-V1.0 offers substantial commercial value(CCICR-V1.0 contract) for widespread adoption in the Chinese cooking domain, enhancing operational efficiency and quality in food preparation.

The remaining content of this paper is arranged as follows: Section 2 introduces the design of CCICR-V1.0. Including hardware components and software architecture. Section 3 reports the experimental results. Section 4 summarizes the entire content of the paper and the future work.

The design philosophy and system architecture of CCICR-V1.0

This section introduces the design principle and functional objective of CCICR-V1.0. Reveal the technical details of CCICR-V1.0 from: 1. Overal design (2.1), 2. Hardware design (2.2), and 3. Software design (2.3).

Overall design of CCICR-V1.0

This study presents an intelligent cooking recorder, which contains a core control board, temperature sensor, pressure sensor, power control module, camera, button, voice module, LCD screen, controller, bus, dish weighing module, and seasoning weighing module. The dish weighing module comprises a support frame, weighing sensor, and dish, while the seasoning weighing module consists of a support frame, weighing sensor, and seasoning container. Both modules can be expanded with multiple units through the bus. The core control board effectively manages and records data from various components, such as the temperature sensor, pressure sensor, power control module, camera, button, voice module, and LCD screen. CCICR-V1.0 has been intergrated into a commercially developed Chinese cooking robot. Due to commercial confidentiality requirements, and after negotitation, this article publicly discloses the collaboration agreement as evidence. The relevant collaboration agreement can be accessed via contract link. The structural topology of CCICR-V1.0 is illustrated in Fig. 1.

Fig. 1
figure 1

The layout sample diagram of CCICR-V1.0.

Full size image

The cooking recorder integrates sensors and monitoring circuits onto standard cooking pots, allowing automatic and precise recording of essential parameters and states during the cooking process through human-machine interaction (via buttons, LCD screens, voice modules, and mobile terminals). The temperature sensor captures real-time temperature changes, while the pressure sensor accurately records ingredient and seasoning weights and their order of addition. Additionally, the camera captures and recognizes ingredient maturity in real-time, storing the information on a SD card. The language interaction module records cooking methods, dish types, weights, corresponding seasoning types, seasoning weights, and other key parameters, as well as pictures and audio-video information during cooking. It also controls the camera, pressure sensor, temperature sensor, voice module, and LCD screen. The system architecture diagram of CCICR-V1.0 is presented in Fig. 2.

Fig. 2
figure 2

The system architecture diagram of CCICR-V1.0.

Full size image

By incorporating sensors and methods, CCICR-V1.0 significantly enhances the automation and intelligence of the product. The primary automation and functions are outlined in Table 1.

Table 1 The main automation and functions of CCICR-V1.0.
Full size table

Hardware design

The hardware system of CCICR-V1.0 is based on STM32F407ZGT6 microcontroller. It mainly includes modules for data storage, human-machine interaction, status indication, data acquisition, video, and data transmission interfaces. The CCICR-V1.0 primarily achieves functions such as data acquisition and storage, including temperature and humidity, pressure, and attitude sensor data. By introducing the FreeRTOS operating system and file system, the system has better real-time performance and more convenient data storage and operation management. The circuit framework diagram of the core board of the data logger is shown in Fig. 3. The core board provides multiple IO ports for expansion to interfaces such as SPI, IIC, FSMC, and USART. And the detailed disgram of sensor modules is presented in Fig. 4.

STM32F407ZGT6 operates at a frequency of 168MHZ. This chip integrates FPU and DSP instructions, and features 192KB SRAM, 1024KB FLASH, 12 16-bit timers, 2 32-bit timers, 2 DMA controllers, 3 SPIs, 3 IICs, 6 serial ports, 2 USBs (supporting HOST/SLAVE), 3 12-bit ADCs, 2 12-bit DACs, 1 camera interface and other functions or interfaces, including 112 general IO ports. STM32F407ZGT6 is widely applied in various embedded systems and IoT devices. CCICR-v1.0 employs the Keil uVision Integrated Development Environment (IDE) and the support package for STM32F4 series to facilitate rapid product development. By referring to the sample code in the STM32F4 standard library, the development efficiency can be improved; the project is compiled to check for syntax errors or other issues; after the program is downloaded to the STM32F4 processor, it can be debugged online with Keil IDE, such as setting breakpoints, stepping execution, and viewing variable values; during the debugging process, oscilloscopes, multimeters, logic analyzers and other tools can be used for assistance to check signal waveforms, voltage, logic status, etc.; based on the debugging results, the program code is modified to optimize performance until the product requirements are met.

Fig. 3
figure 3

Structure of CCICR-V1.0.

Full size image
Fig. 4
figure 4

Detailed diagram of sensor modules.

Full size image

Voice module LD3320

CCICR-V1.0 incorporates the LD332032 speech recognition module, which is based on the ’keyword list’ recognition technology. The speech recognition circuit is illustrated in Fig. 4-U1.It involves the analysis of a substantial amount of speech data (equivalent to tens of thousands of hours of valid voice data collected from thousands of individuals) by linguists to establish a mathematical model. This model is then iteratively trained to extract detailed features of elementary speech units and to discern differences in features between these units. This process yields the optimal statistical probability of various elementary speech features. Subsequently, senior engineers convert the algorithm and speech model into hardware chips and integrate them into the speech recognition module.

By modifying the ’keyword list’ of the speech recognition module, common cooking control phrases can be identified, and the recognition results can be output to the control chip STM32F411. The control chip then manages the cooking equipment and records cooking data based on the speech recognition results, effectively reducing the complexity of human-machine interaction and enhancing the efficiency of cooking data recording.

Weight sensor HX711

HX71133 is a 24-bit A/D converter chip designed specifically for high-precision weight sensors, with advantages such as high integration, fast response, and strong anti-interference ability. The weighing range is 0 to 5 kg, with a weight error of no more than \(\pm 0.005\)kg. The detailed diagram of HX711 is presented in Fig. 4-U2. HX711 communicates with the MCU using 2-wire serial data communication, with timing similar to SPI but not a standard SPI protocol. HX711 is used for detecting the weight of food and seasoning, automatically recording important data such as the weight of ingredients and the amount of seasoning added during the cooking process.

OV2640 camera

OV264034 is a 1/4-inch CMOS UXGA (\(1632\times 1232\)) image sensor produced by Omni Vision company. As depicted in Fig. 4-U3, OV2640 has a small size and low operating voltage, providing all the functions of a single-chip UXGA camera and image processor. Through SCCB bus control, various resolutions of 8/10-bit image data can be output, including whole frame, sub-sampling, scaling, and window retrieval. The UXGA image of this product can reach a maximum of 15 frames per second (SVGA can reach 30 frames, CIF can reach 60 frames). The CCICR-V1.0 camera is utilized to record the real-time cooking behavior of the chef, making it convenient for users to learn through vision. Machine vision can also be employed to recognize the chef’s cooking behavior and automatically record cooking behavior information, further enhancing the automation of information recording.

Six axis sensor MPU6050

MPU605035 is the world’s first integrated 6-axis motion processing component launched by InvenSense company, with a full angular velocity sensing range of \(\pm 250\), \(\pm 500\), \(\pm 1000\), and \(\pm 2000^\circ \)/sec (dps). The circuit diagram of MPU6050 is presented in Fig. 4-U5. It can accurately track fast and slow movements, and its communication interface is IIC interface, used for identifying and recording the speed and frequency of frying during the frying process. By initializing the MPU6050 on the handles of the cooking pot and spatula, CCICR-V1.0 can obtain the 3-axis angular velocity and 3-axis angular acceleration of the cooking utensils. The DMP is used to calculate the Euler angles of the utensils at this time, which is used to obtain motion data for the utensils and spatula for analysis. This allows for intelligent recognition of the stirring speed and frequency during the cooking process, and enables automatic identification and recording.

TFT LCD touchable screen

The TFT LCD touchscreen utilized in this design is the ATK-4.3-inch TFT LCD module, featuring a resolution of \(480\times 800\) pixels. As shown in Fig.4-U6, the driver IC is NT3551036. This IC operates through a 16-bit parallel interface. The capacitive touchscreen, denoted as GT914737, is equipped with an I2C interface. The integration of a touchscreen enhances the intuition and efficiency of operation, allowing for the implementation of visually appealing interfaces for displaying real-time data, status information, charts, and facilitating data recording. Users can visually monitor and control the equipment, make prompt decisions, and easily inspect and rectify the data recorded by CCICR-V1.0.

MLX90614 none contact infrared temperature sensor

Melexis MLX9061438 is an infrared thermometer used for non-contact temperature measurement. Its communication interface is IIC, and the temperature of the measured object ranges from \(-70^{\circ }C\) to \(380^{\circ }C\). Mainly used for non-contact detection of the temperature during the cooking process of frying utensils. The detailed diagram of MLX90614 is presented in Fig. 4-U7.

Software design

This subsection introduces the software design of CCICR-V1.0, including the overall design, the interrelation between different modes, and the software workflow.

Fig. 5
figure 5

System operation flowchart.

Full size image

The design of software flow

The system software’s overall design comprises the STM32 microcontroller as the main control unit, utilizing the C language for sensor data collection, data signal processing, and data storage.

Upon power-up, the STM32 microcontroller initiates a reset operation. Following a successful reset, the microcontroller must verify the hardware connections of three sensor modules. If the hardware circuit connections are abnormal, the sensor circuit connections require rechecking. In the event of normal sensor connections, the pressure sensor, attitude sensor, and temperature sensor need to be initialized, and corresponding operating modes need to be configured. Subsequently, a check is conducted to ascertain the successful connection of the memory card. Upon detecting a successful connection, the memory card is initialized, and upon successful initialization, the system functions are executed. System functionality is managed through four types of human-machine interaction: mobile terminals, buttons, touchscreens, and intelligent voice control. The workflow diagram of the system software is depicted in Fig. 5.

Software function design

The Intelligent Cooking Recorder facilitates the recording of cooking behaviors through four types of human-computer interaction methods, significantly enhancing the convenience and efficiency of data recording. These methods encompass mobile terminals, buttons, touch screens, and smart voice, with the touch screen, mobile terminal, and smart voice interaction methods being particularly convenient. The mobile terminal can establish a connection with the recorder via methods such as the WiFi module and Bluetooth, enabling control of the recorder. The mobile terminal necessitates binding to the recorder through a user account and login password. During the operation of the cooking machine, the cooking recorder must be in a preoperative state, and its sensors must be securely linked to the cooking machine device. The recording of recorder information supports both automatic and manual modes, rendering the cooking machine applicable in a broader range of scenarios. As presented in Fig. 6, the specific operational methods are outlined below:

  1. 1.

    Automatic Mode: The recorder is linked to the cooking machine, enabling interaction with the cooking machine and automatic recording of information such as cooking dishes, seasoning addition time, seasoning names, seasoning types, power curves, cooking pot temperature curves, and dish flipping or rolling (including frequency and duration). It also captures the initial state image of the cooking dish, the end image, and videos of the cooking process.

  2. 2.

    Semi-Automatic Mode: The semi-automatic mode refers to a situation where there is no information exchange between the cooking machine and the recorder. In this mode, users can manually or verbally record the names of the dishes, seasoning ingredients, cooking start time, and cooking end time. However, details such as the number of stir-fries or flips (including frequency and duration), temperature curve of the cooking pot, images of the initial state of the dish, images at the end of cooking, and videos of the cooking process can be automatically recorded.

Fig. 6
figure 6

Diagram of Software Function Design.

Full size image

Experimental implementation

Upon completion of the hardware and software configuration for CCICR-V1.0, experimental validation of the data logger’s performance is conducted. The 3D rendering and the real CCICR-V1.0 core board PCB is shown in Fig. 7. The core board of the CCICR-V1.0 is equipped with abundant peripherals, including the ability to connect an external camera, SD card, WiFi module, voice module, serial port, TFT LCD module, six-axis sensor, and temperature and humidity sensor. This enables real-time detection of seasoning addition amounts, power changes, wok flipping status, temperature, humidity, and the automatic recording of various key data during the cooking process. This work refers to the descriptions of Chinese cooking techniques9, and in conjunction with suggestions from industry partners, the experiment settings are designed as follows: includes testing for weight accuracy, angle accuracy, temperature accuracy, response time and automation rate.

Fig. 7
figure 7

Core board of CCICR-V1.0.

Full size image

Experiment on HX711 weight accuracy

The weight of elements such as salt, spices, MSG, water, and ingredients is an important factor in Chinese cooking9. To this end, CCICR-V1.0 uses the HX711 to record the weight changes of the contents in the smart cooking machine, enabling accurate data collection and material dispensing control functions.

To evaluate the accuracy of weight sensor. The study utilized the Yiheng M1 standard weight set and a 5-standard electronic balance (with a weighing range of 0−500 g and an error of \(0.1\%\)). The weight set comprises weights of 1g, 2g, 5g, 10g, 20g, 50g, 100g, 200g, and 500g. The weighing module of the cooking recorder measures the weight of each standard weight and records the corresponding weight to assess the accuracy of the weighing module. The experimental results are presented in Fig. 8 and Table 2.

Table 2 Weight test data of HX711 weight sensor.
Full size table
Fig. 8
figure 8

Experimental results on the HX711 weight sensor.

Full size image

This work adopts following method to calculate the maximum error (Eq. 1):

$$\begin{aligned} \begin{aligned} \delta _{max} = \mathop {\max }\limits _{\mathrm{{i}} = 1...n} ({\delta _i}) = \mathop {\max }\limits _{\mathrm{{i}} = 1...n} \frac{{{\mathrm{{W}}_{\mathrm{{iSTD}}}} - {\mathrm{{W}}_{\mathrm{{iTest}}}}}}{{{\mathrm{{W}}_{\mathrm{{iSTD}}}}}} \end{aligned} \end{aligned}$$
(1)

The weight test data(Table 2) provides a comprehensive view of the relationship between actual weights and the corresponding average weights recorded by the system, along with their relative errors. The reference weights listed in the ’Weights (g)’ column serve as benchmarks for evaluation, while the ’Average Weight (g)’ column shows the system’s reported weights, demonstrating high precision, especially for lighter weights. This precision is crucial for precise ingredient measurements in cooking applications. The ’Relative Error (%)’ column reflects the percentage error relative to the actual weights, with low values indicating consistent and reliable performance across different loads.

Specifically, spices, salt, monosodium glutamate, and other condiments in Chinese cooking are often used in single servings of less than 10 grams. Water, oil, and cornstarch are commonly measured in the range of 10 to 100 grams. Other ingredients, such as vegetables and meat, are typically measured in the range of 100 to 500 grams. This weight accuracy experiment covers almost the entire range of commonly used weights in Chinese cooking, meeting the technical specifications for Chinese culinary practices.

Based on the data presented in Fig. 8 and Table 2, it is evident that the HX711 weight sensor exhibits a high degree of accuracy and reliability. Analysis of the error magnification reveals that although there are anomalous points in the second, third, fourth, and fifth experiments when the error is amplified by 50 times, the average error of all detection points is only 0.023. Furthermore, the maximum deviation (\(\delta _{max}\)) is calculated to be only 0.288%, indicating a close match between the average weight measurements and the actual weights.

Overall, the data underscores the system’s accuracy and reliability, which are essential for achieving consistent cooking results. The HX711 weight sensor demonstrates great precision and is well-suited for applications that require precise weight measurements, particularly in the context of cooking.

Experiment on MPU6050 angle accuracy

This section of the experiment aims to measure the accuracy of the MPU6050 sensor. The angle sensor accuracy of the CCICR-V1.0 is tested on a standardized measurement platform. To ensure consistency in the experiment, the sensor is integrated onto a development board and placed on the angle measurement platform, where pitch, roll, and yaw angles are measured at various angles. The angle range is set from -90 to 90 degrees to simulate the specific movements characteristic of Chinese cooking techniques. The schematic diagram of the experimental setup is shown in Fig. 9, and the experimental results are detailed in Table 3.

Fig. 9
figure 9

The diagram of MPU6050 sensor experiment deployment.

Full size image
Table 3 Angle test data of MPU6050 sensor.
Full size table

As presneted in Table 3, the MPU6050 sensor performs well in tests at different angles, with measurements for Pitch, Roll, and Yaw closely aligning with the actual angles and showing relatively small deviations. Specifically, Pitch ranges from −89.82\( ^\circ \) to 90.29\( ^\circ \) and Roll from −89.74\( ^\circ \) to 90.29\( ^\circ \), demonstrating consistent accuracy. However, Yaw exhibits slightly larger deviations at extreme angles (−90\( ^\circ \) and 90\( ^\circ \)), the average error is 0.22\( ^\circ \), possibly influenced by specific factors. In practical applications, these minor precision deviations may impact scenarios that require accurate measurements, so further calibration may be necessary to enhance the sensor’s measurement accuracy for specific uses.

Experiment on MLX90614 temperature accuracy

The temperature experiment is set according to the cooking temperatures of Chinese cuisine9. The experimental testing temperatures cover those used in techniques such as fermentation, pasteurization, drying, frying, boiling, steaming, baking, and making soups in Chinese cooking. The specific experimental steps are outlined as follows:

  1. 1.

    Arrange the heat source, the MLX90614 non-contact temperature measurement module, and the DT1311 contact temperature measurement device. The K-type thermocouple probe of the DT1311 should be in direct contact with the cooking pot, while the MLX90614 module should be positioned 1 meter away, with its sensor aligned with that of the DT1311.

  2. 2.

    Activate the temperature controller of the heat source and set the temperature to \(257^\circ {C}\). Record the start time of the experiment and the initial temperature of the heat source.

  3. 3.

    Record the heat source temperature and the sensor-measured temperature at 1-minute intervals.

  4. 4.

    Conclude the experiment when the heat source temperature reaches \(257^\circ {C}\).

To minimize measurement errors from the MLX90614 sensor and enhance temperature accuracy, an iterative compensation model is established. This model accounts for changes in the sensor’s measurement error, as illustrated in Eq. 2.

$$\begin{aligned} \begin{aligned} {T'_{i}} = \left\{ {\begin{array}{*{20}{l}} {{T'} \pm 2\varepsilon }& {\left| {\Delta {T_t}} \right| > \varepsilon }\\ {{T'} \pm \varepsilon }& {\varepsilon /2< \left| {\Delta {T_t}} \right| \le \varepsilon }\\ {{T'} \pm \varepsilon /2}& {\varepsilon /4< \left| {\Delta {T_t}} \right| \le \varepsilon /2}\\ {{T'}}& {0 < \left| {\Delta {T_t}} \right| \le \varepsilon /4} \end{array}} \right. \end{aligned} \end{aligned}$$
(2)

Here, \(T'( ^{\circ }C)\) represents the actual temperature of sensor. \({T'_{i}}( ^{\circ }C)\) represents the measured temperature value of the sensor after the i-times compensation; \(\varepsilon ( ^{\circ }C)\)is the system error. \(\Delta {T_{t}}\) is the difference between the temperature measurement data of MLX90614 and DT1311 at time t, when both indicators are less than \(\varepsilon {/4}\), the iteration ends.

The average error between the non-contact temperature measurement value of MLX90614 and the temperature measurement value of the contact thermometer DT1311 belongs to the system error, which is presented in Eq. 3:

$$\begin{aligned} \begin{aligned} \varepsilon {{ = }}\sum \limits _{t = 1}^n {\sqrt{\frac{{{{(\Delta {T_t})}^2}}}{n}} } \end{aligned} \end{aligned}$$
(3)

Where \(\varepsilon ( ^{\circ }C)\) represents the system error, n means the heat temperature rasising steps, \(\Delta {T_t}\) means the temperature error of MLX90614 and DT1311 at time t.

The error corrected MLX90614 non-contact temperature measurement value and the temperature measurement value, temperature error, and relative error of the contact thermometer DT1311 are shown in Fig. 10 and Table 4.

Table 4 Temperature detection error of MLX90614 sensor(Temp is short for temperature, Avg-T is short for average temperature, Avg-E is short for average error).
Full size table

Table 4 presents the detection errors of the MLX90614 sensor at different temperatures. Analysis of the data indicates that differences at specific temperatures (Temp) significantly influence the measurement accuracy of the sensor. In the lower temperature range (below \(25^\circ \textrm{C}\)), as the temperature increases, the detection error tends to rise; for instance, at \(26^{\circ } C\), the error is \(0.23^\circ C\), while at \(40^\circ C\), it increases to \(0.134^\circ \textrm{C}\). This may relate to the performance characteristics of the sensor under low-temperature conditions. In the higher temperature range (above \(100^\circ \textrm{C}\)), the detection error shows some fluctuations. For example, at \(161.49^\circ \textrm{C}\), the error is \(-0.164^\circ \textrm{C}\), and at \(257.05^\circ \textrm{C}\), it is \(-0.168^\circ \textrm{C}\). These fluctuations may reflect stability issues of the sensor at high temperatures. Therefore, to ensure measurement accuracy in practical applications, temperature calibration or compensation algorithms should be implemented. Additionally, factors such as external interference and sensor aging need to be considered to improve measurement precision and stability across different temperatures.

Fig. 10
figure 10

Experimental results on the MLX90614 sensor.

Full size image

To improve the presentation of data distribution, we normalized the data from the same location and increased its magnification to 50 times. As illustrated in Fig. 10, the maximum absolute error for the MLX90614 non-contact temperature measurement module, employed in CCICR-V1.0, at point 19 is \({0.258}^{\circ }C\), and the maximum relative error is \(0.88\%\) and point 0. After amplifying the error by 50 times, it unveils abnormal points in steps 1, 3, 5, 6, 7, 10, 16, and 28. However, a closer examination of the temperature variation curve in the graph and Table 4 reveals that the error at all detection points is merely -0.0321. These experimental results confirm the high-precision temperature detection capability of CCICR-V1.0 and its compliance with intended functionalities.

Moreover, the infrared temperature measurement module integrated in this design has an approximate cost of 200 yuan, which is only 1/12 of the price of SA50AGW, a high-precision digital thermometer (\({0-500}^{\circ }C\), \(\pm 1\%\) or \({\pm 1}^{\circ }C\), sold at 3600 yuan) from a reputable domestic brand named Shiao Technology. The module offers superior accuracy, better cost-effectiveness, and holds considerable commercial potential for promotion.

CCICR-V1.0 response time and automation rate measurement

The STM32F407-based system operates with a working frequency of 168MHZ, ensuring efficient processing capabilities. By utilizing the processor timer with its counting function, this system accurately measures and acquires data related to pitch, roll, and yaw angles. The conversion of these angle measurements into readable data is swiftly accomplished with a system response time ranging between 1683-1684 microseconds. Additionally, the system seamlessly displays the information related to pitch, roll, and yaw angles on a TFT LCD screen for convenient visualization.

The response time results is presented in Table 5. In Table 5, time variables are defined as time-1, time-2, and time-3. Time-1 is the timer count value when the MPU6050 begins to detect the pitch, roll, and yaw angles; time-2 is the count value at the completion of the detection. By calculating the difference between time-2 and time-1, the time required for the MPU6050 to complete the hardware conversion of pitch, roll, and yaw angles can be obtained. Time3 records the count value when reading the data of these three angles and displaying the corresponding values on the LCD screen. This function is based on the successful completion of the detection of the three angles and resets the timer counter after the data is displayed. Finally, the difference between time-3 and time-2 is used to calculate the time required to display pitch, roll, and yaw in three lines. This process demonstrates the effective integration of the MPU6050 sensor’s functionality and data output. The tests reveal that displaying three lines of data on the LCD takes between 7800-7801 microseconds, which translates to an average display time of approximately 2600 microseconds per line. This level of performance is further complemented by other key features like the HX711 pressure sensor and MLX9064 temperature sensor, which convert and acquire data within 3 milliseconds.

Table 5 Experimental results on time response of CCICR-V1.0.
Full size table

The operation of the cooking recorder encompasses recorder startup \((R_{s})\), recording cooking dishes \((R_{Dishes})\), seasoning addition time \((Rt_{Flavour})\), seasoning names \((Rn_{Flavour})\), seasoning weight \((Rw_{Flavour})\), power curve \((Rc_{Power})\), temperature curve \((Rc_{Temp})\), dish flipping or rolling (including frequency and duration) \((Rs_{Stir-fry})\), video recording \((R_{Video})\), and other behaviors. By analyzing the behaviors or steps in the cooking process, the automation rate \((R_{Auto})\) of the recorder can be calculated. Table 2 presents the automation rate analysis table of the cooking recorder in this design and the manual recording of traditional cooking. In the table, Y represents automatic, and N represents \(non-automatic\).

Table 6 Table of CCICR-V1.0 automation rate.
Full size table

We adopt following method to calculate the Automation Rate of CCICR-V1.0 (Eq. 4):

$$\begin{aligned} \begin{aligned} {R_{Atuo}} = \sum \limits _{i = 1}^n {\frac{{{R_{iAuto}} = Y}}{{{R_{iAuto}}}}} \end{aligned} \end{aligned}$$
(4)

Based on the data from Table 6, the Automation rate is calculated as \({R_{Atuo}} = 0.889\). This indicates that the intelligent cooking recorder exhibits a high level of automation. In comparison with the manual recording of traditional cooking, the automation rate has significantly increased from 0 to \(89.9\%\), effectively reducing the workload and error rate associated with manually recording various data in the cooking process. This provides quality assurance for the program control of intelligent cooking machines.

Conclusions

The Chinese Cuisine Intelligent Cooker Recorder-V1.0 (CCICR-V1.0) revolutionizes data collection in Chinese cooking by addressing the challenges posed by intricate and diverse cooking processes. Employing posture, temperature, and pressure sensors, CCICR-V1.0 automatically records critical data during cooking, boosting automation efficiency to 89.9%. This innovative device leverages multitasking strategies to ensure efficient high-frequency data acquisition across multiple channels in complex environments. The storage method effectively allocates CPU resources, ensuring real-time storage on large-capacity devices such as NAND Flash for secure storage, management, and preservation of high-quality data. The precision of CCICR-V1.0 weighing module is unparalleled with a maximum relative error of 0.288%. Furthermore, non-contact infrared temperature measurement module demonstrates exceptional accuracy with a maximum absolute error of \({0.258}^{\circ }C\) and a maximum relative error of 0.88%. Notably, this measurement module achieves a cost reduction of approximately 91.7% while maintaining an accuracy improvement of around 40% compared to similar products in the market. Additionally, CCICR-V1.0 real-time data recording capabilities enable automatic recording of essential parameters throughout the cooking process, enhancing recording efficiency and accuracy. In summary, CCICR-V1.0 holds immense commercial value for widespread adoption in intelligent cooking, owing to its ability to effectively capture data related to Chinese cooking processes. In the future, CCICRs will evolve through product iterations driven by market feedback. This will enhance functionality, incorporate more IoT features, and further reduce costs.