The influence of recording conditions on the results of acoustic analysis

The first test was to examine the influence of noise and distortions on the complicated process of analyzing recordings.

Examination methodology

In the first stage, model samples of male and female voices (sample of extended phonation of the "a") were recorded under conditions considered optimal: using a Behringer B1 condenser studio microphone with a dedicated microphone preamplifier and a music card (the so-called M-Audio audio interface) MobilePRE , in a soundproofed and electromagnetically shielded anechoic chamber, using a laptop running on battery power and enclosed in a metal container. Then the disturbing signals were recorded:

• various sounds (knocks, conversation, crackles, scrapes) – noises from hallway,
• mains noise,
• vibrations and impacts (knocking on the housing of the microphone),

and files with white noise were prepared. Using a specially written computer program (Fig. 1), the signals were combined in a strictly defined energy ratio, so as to generate a signal with the appropriate interference at a distance of 10, 20, 30 or 40 dB. In addition, the test samples were generated without interference, but inverted in phase, as well as the samples distorted by 3, 5, 10 dB.

The samples were loaded into our test software and analyzed. The acoustic analysis was performed automatically without operator intervention. The results were checked for the occurrence of the so-called gross errors in the determination of the fundamental frequency, however, no error occurred.

Fig. 1. A window of a specially prepared program for combining test waveforms

Results

Figure 2 shows sample waveforms generated during the tests together with analyzes results. There is a clear increase in the value of the parameters already in the case of the distance from noise at the level of 30 dB.

a)

b)

a) reference waveform

a)

b)

b) waveform with white noise – noise interval 40dB

a)

b)

c) waveform with white noise – noise interval 30dB

Fig. 3 shows graphically the influence of various disturbances on the determined values of voice acoustic analysis parameters. The tested types of disturbances are placed on the X axis:

• inw – signal inversion,
• clip – limitation caused by distortion,
• przy – mains hum (interference from the lighting network),
• szum – white noise,
• wib – vibration and impact, finger tapping on the microphone housing,
• rozn – various sounds from the environment (knocking, talking, crackling, scraping).

The number after the name of the disturbance type means in the case of distortion - the amplitude degree of distortion expressed in dB, and in other cases the signal-to-disturbance ratio expressed in dB. The green color means that there is no influence of disturbances on the parameter values, the blue color - the reduction of the determined parameters, and the red color - the increase of their values.

The tests performed showed the greatest influence of ambient noise. The full correctness of the test is ensured only by the difference of 40 dB between the recorded sound signal and the disturbance. A similar situation occurs for mains hum, while the influence of vibration and signal reversal is negligible. The effect of distortion is relatively small, but since it lowers the values of parameters, they should be absolutely eliminated at the recording stage.

Fig 3. Comparison of the influence of disturbances and distortions on the results of acoustic analyzes: red - overstatement of the analysis results, blue - understatement