Design of a Signal Conditioning Device for Remote Breath and Swallowing Sounds Recording

Biological signal conditioning devices, such as Biopac amplifiers, are designed to condition a variety of biological signals (i.e. EMG, ECG, breath sounds, etc.). The bandwidth of these signal conditioning devices is usually between 0.5 Hz and 5KHz. In studies that need the simultaneous recording of both a high frequency signal (i.e. tracheal breath sounds) and a very low frequency signal (i.e. respiratory airflow signal), two different signal conditioning device must be used in order to provide the bandwidth of interest for each signal. In clinical application however, it is desirable to have one small device including hardware and software to achieve recording and analyzing of all signals of interest. One particular application is to have an integrated device for recording breath and swallowing sounds along with a respiratory airflow signal. The goal of this work was to design such a device that also had simple adjustment of the breath or swallowing sounds' level and the flexibility for multiple recordings with different file names given by the user.Breath sound signals contain information about the physiological characteristics of the airway and have been used in apnea/hypopnea detection [1]. Also, breath and swallowing sounds analysis can be used to identify individuals with swallowing disorders [2–4]. The respiratory airflow, that is a very low frequency signal, is also desired to be recorded simultaneously with breathing sounds for monitoring sleep apnea disorder. Our group has been involved in developing automated swallowing disorder detection algorithms that require hardware to capture the necessary data.Aside from the fact that commercial audio recorders cannot record DC or low frequency signals, there are also other limitations. For example, a simple handheld recorder (e.g. TASCAM DR-07 mkII) is insufficient because the best signals are obtained when a microphone is stationary over the trachea. Stand-alone desktop recorders (e.g. Edirol R-44) can accept external microphones but they have built in (not adjustable) filters. In addition, assigning custom names (e.g. patient_date) to different recordings are not easy as these devices assign automated non-descriptive numeric names to the recorded files. A standard data logger, such as those offered by National Instruments (e.g NI USB-6008) is capable of recording these signals with some minor analog interfacing. However, such products are tied to computer software, and such software is not often available in a clinical application. This work looks to resolve these common issues in one standalone device.The final design features two main parts. The first part is the recording box that connects with the microphones and pneumotachograph, and contains a removable SD card to store data. The recording box holds the analog signal conditioning circuit and the microprocessor that controls data acquisition. The second part is an Android-based Toshiba tablet that runs a custom app to control recording process and filing wirelessly.The primary purpose of the analog circuit is filtering and amplification for three channels (two audio, one airflow). Both audio channels feature a low noise instrumentation amplifier with adjustable gain, followed by Sallen-Key high pass and low pass filters. The gain for the audio channels is adjustable between 20 to 60 dB, and for airflow between 34 to 79 dB. The frequency bandwidths were designed to be 20 Hz-5 kHz, and 0.05 Hz-2.5 kHz for the audio and airflow channels, respectively. Each channel also contains a separate high speed 16-bit analog to digital converter. These converters are triggered and read by the controlling microprocessor at 10 kSPS. The device also has a mixer and headphone amplifier so that the user can listen live to the audio channels to properly adjust the gain of the sound recording. For additional monitoring, an LED bar array has been included to give a visual indication of the signal level.Data acquisition, writing to SD memory, data processing, and communication with the tablet are all carried out by an Arduino Mega ADK microcontroller board. The board is based on a 16 MHz ATmega2560 processor. The Arduino Mega was selected for several reasons. It is programmed in C programming language, has a variety of ports and digital I/O, and is easy to interface with other hardware. The Arduino reads in the 16 bit data from the analog to digital converters serially over separate the serial peripheral interface (SPI) buses. After the data is acquired, it is written over a separate SPI bus to an SD card. Data is logged to a binary file, which can be converted to a standard text file.The user operates the device through two main control interfaces. The first is on the front panel of the recording box. Channel gain knobs, monitor level indicators, and a variety of indicator LEDs exist here. Figure 1 shows the device's front and rear views. The second user interface (Fig. 2) provides buttons to select channels to record, start/stop recording, name the recorded files, manage and convert files contained on the SD card. The control of recording, file naming, and channel selection is done via the Android-based tablet (connected to the Arduino via Bluetooth) running the custom app.We evaluated the device by comparing the signal and noise records compared with Edirol R-44 audio recorder. We recorded a series of pure tones, ranging from below audible frequency to above 5 kHz. The pure tones were played through a pair of speakers (Cyber Acoustics desktop speaker system, CA-2100), and recorded by both devices simultaneously. In the second test, the breathing sounds of a healthy individual were recorded. The participant was asked to breathe at different flow rates starting at very low flow rate and slowly increase the flow rate to a maximum, and then slowly decrease the flow rate. The second test could not be performed by both recorders simultaneously. However, both tests were performed on the same individual with identical microphone positioning, in immediate succession. The spectra of the two recordings were compared in terms of average power of the breath sounds over different frequency bands.The device has been used by several individuals other than the designer, and it has operated as designed. Minor software revisions were done during the design process in response to user feedback.Figure 3 shows the frequency content of two recorded signals, one from the new device, and one from the benchmark R-44 over the recording time. The dark colouring (red) represents the strongest frequency component, and the light colouring (white and blue) shows the weakest components. The red squares are the recorded pure tones, and the green components around the pure tones represent undesirable noise in the signal around the tones. Clearly, the new device records less noise than the R-44, though not by a significant margin. The noise level of the R-44 is acceptable for obtaining accurate results, so any improvement just makes analysis more reliable. In some recordings, there is a noise component above 3 kHz that is a known imperfection in our device. The source of this slowly varying noise component is likely a thermal drift in the analog to digital converters. However, its presence does not adversely affect the analysis as the signals of interest (breath and swallowing sounds) have their major components below 3 kHz.Apart from the attributes of the recorded signals, the rest of the design met the goals. It can log 16-bit error free data from all 3 channels simultaneously at 10 kSPS to the SD memory card. Even though it requires custom software to carry out recording, it is streamlined for this purpose, and works on the highly portable tablet platform. The tablet interface works well and provides a smooth user experience.Our designed device achieves the technical functionality of its counterparts, with a simplified user interface. The custom design of the device makes it an appropriate tool in research and clinical applications.This project was supported by the National Science and Engineering Research Council (NSERC) of Canada, and the Faculty of Engineering of the University of Manitoba.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.