Project Scope:
Nocturnal asthma is a widespread and costly disorder that is currently very difficult to
accurately and conveniently monitor. Given the particular prevalence and negative effects of NA
in children, we propose to design a commercial, home-based device capable of monitoring
symptoms and alerting parents or caregivers if intervention may be required (i.e., during an
asthma attack). The product will not be diagnostic for asthma, and is intended for children who
have already been diagnosed with asthma and may suffer from NA. Our device will also not
disrupt the child as she/he sleeps, and will be easy to configure and use since its target
audience is not medically trained. Given the desire for a low-cost, home-based, autonomous
device, we plan to use audio signals, at least in part, to monitor an asthmatic child’s nocturnal
coughs.
Nocturnal asthma is a widespread and costly disorder that is currently very difficult to
accurately and conveniently monitor. Given the particular prevalence and negative effects of NA
in children, we propose to design a commercial, home-based device capable of monitoring
symptoms and alerting parents or caregivers if intervention may be required (i.e., during an
asthma attack). The product will not be diagnostic for asthma, and is intended for children who
have already been diagnosed with asthma and may suffer from NA. Our device will also not
disrupt the child as she/he sleeps, and will be easy to configure and use since its target
audience is not medically trained. Given the desire for a low-cost, home-based, autonomous
device, we plan to use audio signals, at least in part, to monitor an asthmatic child’s nocturnal
coughs.
Design Specifications:
Cost: must cost less than $100
Size: Must have width and depth less than 12 inches by 12 inches. Must have a height
less than 24 inches.
Weight: Less than 8 pounds
Accuracy: Must recognize at least 70% coughs. Can falsely recognize up to 2 coughs
per hour
Recording time: 12 hour minimum
Sampling rate: Appropriate for chosen modality such as;
Maximum audio alert volume: 80 dB
Transmitter range: An open field range of at least 1,500 feet
Operating noise: Must be less than or equal to 30 dB
Other Considerations:
Intended for children of ages 5-18
The device is not diagnostic
Must not harm the child being monitored and must not disturb the child’s sleep
Must recognize coughs in real-time or near real-time
Must not constrain the child
Must have a minimal, easy setup comparable to a baby monitor.
Cost: must cost less than $100
Size: Must have width and depth less than 12 inches by 12 inches. Must have a height
less than 24 inches.
Weight: Less than 8 pounds
Accuracy: Must recognize at least 70% coughs. Can falsely recognize up to 2 coughs
per hour
Recording time: 12 hour minimum
Sampling rate: Appropriate for chosen modality such as;
- Audio spectra: At least 48 kHz
- EMG: At least 2 kHz22
- EEG: At least 4 kHz23
Maximum audio alert volume: 80 dB
Transmitter range: An open field range of at least 1,500 feet
Operating noise: Must be less than or equal to 30 dB
Other Considerations:
Intended for children of ages 5-18
The device is not diagnostic
Must not harm the child being monitored and must not disturb the child’s sleep
Must recognize coughs in real-time or near real-time
Must not constrain the child
Must have a minimal, easy setup comparable to a baby monitor.
Preliminary Analysis:
Sample calculation:
At 1kHz tone at a distance of 5ft: dB/km = 4.7 attenuation per length (retrieved from chart) dB = (4.7dB/km) * (1km/1000m) * (1m/3.28ft) *( 5ft) dB = 0.0072 attenuation Higher frequency signals are attenuated more than the lower frequencies. Higher frequencies are attenuated more with increasing distance than lower frequencies. These values were measured at 50% relative humidity. On the left column, four time-series plots of a cough: the word “hello”, the word “cheek”, and a clap are aligned by their starting points (first large deflection from baseline). To capture the maximum frequency possible frequency range in Matlab, the sampling rate was set to 96,000 samples/sec. On the right column: the sound spectrograms computed using the STFTs of the signals are shown. For each spectrogram, the window size is 256, nfft is 256, and the overlap is 128. Note that the spectrogram of a cough is very different from the other tested sounds, though the time-series of the sounds are more difficult to distinguish.
On left, the time-series of three coughs with increasing relative loudness as determined by a listener next to the microphone. The microphone and listener were both 5 feet away from the cougher (the width of a twin size bed). On right, the spectrogram of this recording. Note the similarity of coughs when viewed on the right plot and the differences in the signals on the left plot.
|