Impact of the GSM Mobile Phone Network on the Speech Signal – Some Preliminary Findings

Abstract

The landline and mobile phone networks are very different in the manner in which they handle and process the speech signal, with consequential differences in their impact on the speech spectrum as well as on the acoustic parameters of interest in the forensic arena. This paper overviews the key features of both networks which give rise to these differences. Some findings are then presented showing the impact of the mobile network on fundamental frequency, F0, as well as on the first three formants in vowel sounds. It is shown that though the impact on F0 is not great, the impact on the formant frequencies, and particularly on the measurements of these determined using automated formant trackers, can be significant. These early findings are considered to be of great importance to forensic practitioners relying on acoustic measurements as part of their investigations.

Effects of telephone transmission on the performance of formant-trajectory-based forensic voice comparison – Female voices

“…Frequencies far enough outside this band are lost, and frequencies near the edges of the band are distorted. Digital landline-telephone systems sample the signal at 8 kHz and apply relatively lossless compression and decompression (Guillemin and Watson, 2008). Comparing formant measurements made on direct-microphone recordings and landline-transmitted recordings, Kü nzel (2001) and Lawrence et al (2008) found that differences in F2 (and in F3 for the latter study) were generally negligible.…”

“…On the basis of these results, it appears that, given a direct-microphone vs. landline-transmission condition, the use of F1 measurements from tokens of vowel phonemes with intrinsically low F1 should be avoided, but F1 measurements from tokens of other vowel phonemes and F2 and F3 measurements would generally not be expected to present great difficulties (see discussion in Nolan, 2002;Kü nzel, 2002, andRose, 2003 §99.880). Guillemin and Watson (2008) describe a typical mobiletelephone system (Groupe Spécial Mobile/Global System Mobile Communication, GSM, network, including the Adaptive Multi-Rate, AMR, codec) and contrast it with landline-telephone systems. Such mobile telephone systems apply substantial amounts of data compression to the speech signal, and the degree of data compression applied can change from moment to moment (up to once every 40 ms).…”

An investigation of supervector regression for forensic voice comparison on small data

“…The effect of the mobile-telephone system on the original speech signal is often associated with data compression which can vary from moment to moment and can result in different transmission rates in the range of 4.75 to 12.20 kbits/s. The mobile-telephone system has a bandpass filter with a lower limit of 100 Hz to an upper limit of up to 3.6 kHz, whereas a landlinetelephone system has a bandpass filter in the range of 300 to 3,400 Hz together with lossless compression and decompression algorithms with 64 kbits/s transmission rate [50].…”

Impact of Background Noise in Mobile Phone Networks on Forensic Voice Comparison

“…We view this as a much better approach and was thus used in our experiments. The rationale for this approach is that in a mobile phone network the speech codec is the only component responsible for changes that might occur to the speech signal during transmission [5,6]. These codecs have many modes of operation which get selected dynamically in response to events happening in the network as a whole or in response to changing speech characteristics.…”