Acoustic Data Support at LISHA's IoT Platform

SmartData containing raw sound samples, preserving all features, can be sent to the platform, using the Digital Unit of type 2, subtype 35 (as described here). The processing applied to the data is based on specific microphone's specifications.
The WAV files that are sent to the platform using 32 bits per sample encoding, are compressed/inflated on-the-fly at the platform. The WAV file's header is exposed on the table below, according to this reference.

WAV Header

To save storage capacity, once the microphone's characteristics do not demand more than 16 bits to encode sensed data, the WAV data received by the platform are encoded with 16 bits per sample, and then compressed using the Free Lossless Audio Codec FLAC . When data is retrieved from the platform, the data is read from the database, and restored to the original format of 32 bits per sample WAV. The ffmpeg tool is used to compress / restore the data.

SmartData and Time Series

The sound data (the value field of the object) sent to the platform using SmartData in JSON format has to be encoded in Base64 format. The data returned by a get() command will also be encoded this way.

An exemplo of JSON that should be sent to insert a WAV data into the platform, using put.php

'smartdata' : 
        [
            {
                'version' : '1.1',
                'unit'    : 0x02230000,
                'value'   :  <base64 encoded data>,
                'error'   : 0,
                'confidence' : 0,
                "x" : 123123123,
                "y" : 123112233,
                "z" : 83728372,
                't' : 1509391500000000,
                'dev' : 1 
            }
        ]

In order to retrieve the WAV data, use the get.php, providing the same unit, coordinates and device:

'series' :   {
            'version' : '1.1',
	    'unit' : 0x02230000,
            'x' : 123123123,
            'y' : 123112233,
            'z' : 83728372,
	    'r' : 0,
            't0' : 1509390000000000,
            't1' : 1509392000000000,
            'workflow' : 0,
            'dev' : 1  
}

Remarks and Evaluation

The pyflac library and the FFMpeg library (a wrapper for the ffmpeg tool) were also evaluated. The first also demand a call to an external script, and the latter has some broken dependencies (from older PHP versions), and needs files as input. Therefore, no advantages from calling directly the external ffmpeg tool are evident, and no pre-processing (rescaling) is needed before applying compression, as ffmpeg make all necessary processing.

The evaluation was made with 8 sample files. After compressing and restoring each file, the maximum difference between the original data and the restored data was measured. The Peak Signal to Noise Ratio (PSNR) was also calculated for each restored file, compared to the original data. As shown in the table below, the FFMpeg and pyflac shown very similar values for the PSNR results. Therefore, the FFMpeg is employed in this implementation of the platform compression/decompression tool. All files have 320046 bytes size, and the compression shrink them to an average size of 136150 bytes (42,54% of the original size), achieving a good compression ratio.