Wireless communication terminals

[0001] This invention relates to wireless communication terminals, especially mobile telephones, and the hands-free activation of such terminals.

[0002] It is known to incorporate voice recognition software in mobile telephones to allow users to dial a caller by name. However, in order to make use of this facility, the telephone has to be operated manually because, even when in the standby mode, the audio system is not normally turned on. Instead, the receiver only is powered up to receive the paging channel to check for incoming call requests, and for reasons of power saving, the audio system remains turned off.

[0003] According to the invention, a wireless communication terminal is adapted so that it is capable of recognizing a predetermined sound in the vicinity of the terminal and its audio input system is powered on periodically when the terminal is in the standby mode and serves to activate the terminal if said predetermined sound is recognized.

[0004] Preferably, the audio input system is powered up with the paging channel, and preferably only operates during the paging channel for reasons of power saving, and then processes the received audio signal to recognize said predetermined sound if it is present. In a DSP based GSM terminal, the same DSP processor is used for the radio modem and audio processing, and therefore powering up the processor for paging will automatically make the audio processing function available and produce said audio signal if the audio input system is also powered up.

[0005] The paging channel in a mobile telephone consists of a number of paging blocks of short duration separated by an interval of 0.5 to 2.5 seconds. For example, a GSM terminal has a paging channel of four data blocks or bursts, each 4.615 ms long. Each burst has a portion allocated to radio modem processing and the remainder allocated to audio processing, which over four bursts might total 16 ms. Thus, the audio input system of a GSM terminal according to the invention has to recognize said predetermined sound over a short interval of about 16 ms, which would be difficult for a speech pattern. Preferably, therefore, the sound selected is a whistle, which has a narrow bandwidth characteristic and changes only slowly with time so that it can be more easily recognized from a short sample. Also, a whistle can be more easily distinguished from other sounds and will therefore avoid false responses.

[0006] The invention is therefore based on the fact that sound recognition is a useful function that can be switched on periodically in a mobile telephone during the standby mode, either with the paging channel or any other short duration channel such as a monitoring channel, and can then be used to recognize narrow bandwidth sounds such as a whistle, to activate the telephone. Once activated, the telephone may then be responsive to voice commands and may then support a speaker phone mode of application.

[0011] A typical GSM mobile terminal, as illustrated in FIG. 1, comprises a radio module 1 for receiving and transmitting radio signals in respective receiving and transmitting paths RX, TX, a modem 2 to process the signals in the receive and transmit paths, a channel coder 3 to process signals in transmit and receive channels and a speech coder 4 to process speech signals which are either output to a speaker module 5 or received from a microphone module 6. It will be appreciated that the modem 2, channel coder 3 and speech coder 4 are normally incorporated in one digital signal processor DSP, and a rechargeable battery power unit 7 supplies power to all of the above components.

[0012] When such a GSM mobile terminal is in the standby mode, the power unit 7 only powers up the radio module 1 and DSP on a low duty cycle to receive a paging radio channel to check whether an incoming call is being requested. The speaker module 5 and microphone module 6 are not powered up in the standby mode in order to save power until such time as they may be required.

[0013] The paging channel in GSM consists of four data frames or bursts, each 4.615 ms long, as shown in FIG. 2. The DSP is therefore powered up for about 18.5 ms, and this is repeated at an interval of 2.1 seconds. During each burst, the DSP is only processing data relating to the radio modem function, and this only occupies a minor part of the burst, the remaining major part of the burst being reserved for audio processing when the terminal is in call. The total reserved time for audio processing between four bursts totals about 16 ms, and it is a feature of the invention, that this reserved audio processing time is used by powering up the microphone module 6 during this time so that the audio input it generates is processed and compared with a predetermined audio input which is indicative of a “wake up” command from the user.

[0014] Said predetermined audio input is preferably a whistle, this having a narrow bandwidth characteristic which makes it more easily recognizable from a short sample, as illustrated in FIG. 3. The graph shows typical power spectra for both a whistle and normal speech, and illustrates the fact that a whistle is essentially a fairly pure single audio tone, whereas speech contains significant power in more bands across the range. Thus, whistles can be detected from only a short time period because they are easily distinguished from other sounds such as background acoustic noise, which has no sharp peaks, speech which has multiple “formant” frequencies, and music, which like speech has multiple frequencies present.

[0015] It is not necessary that the whistle is of a particular pitch or even that the pitch is held constant with time. The recognition algorithm would merely take a snapshot of the signal and look for a single narrow-band peak much higher than the surrounding signal at other frequencies.

[0016] The key feature of the whistle is that it is narrow-band at all times; it is therefore not necessary to scan for it continuously in order to detect it. The GSM paging cycle allowing 16 ms samples of speech at a maximum of 2.1 s intervals is therefore sufficient for whistle recognition.

[0017] In a simple implementation, it would be necessary for the user to keep whistling for this maximum interval of 2.1 s to ensure that at least one block of audio samples is captured. However, if it turns out that this is too long to maintain a whistle, then the whistle length could be reduced with an increase in power consumption.

[0018] A suitable whistle recognition algorithm needs to detect a narrow-band signal of unknown frequency in the presence of speech with low false alarm probability. A pre-shaping filter would be provided to remove low frequency components from the signal which would otherwise affect the recognition process.

[0019] Reasonable recognition/false alarm results have been obtained using the following algorithm:—

• [0020] (i) If the energy of the block of audio samples is above a threshold then take the FFT for 128 samples sampled at 8 kHz;
• [0021] (ii) find the largest energy bin and find the width of the peak to half the peak power;
• [0022] (iii) find the next largest peak excluding the interval found in (ii);
• [0023] (iv) if the ratio of the energy in the first peak of the second peak is >10 dB then declare that the whistle has been recognized.

[0024] An alternative non-linear approach is based on the low variance of the phase increment per sample in the audio block for a whistle compared with speech.

[0025] Although the algorithm has been discussed in terms of GSM, it will be appreciated that it can be generalized for any wireless communications system. The only requirement is the capability to periodically switch on the audio hardware to sample 16 ms of audio data. All mobile phone systems should fulfill this requirement since the mobile will need to switch itself on periodically either to listen for paging signals (or their equivalent) or for network measurements, and being a phone it should have the appropriate audio capabilities. As long as this duty cycle is sufficient, the algorithm need not be modified.

[0026] In one embodiment of the invention, a mobile terminal is further adapted to include voice dialling and speaker phone operation. The user is then able to use the terminal in hands-free mode as follows:

• [0027] (i) user whistles;
• [0028] (ii) terminal responds with an acknowledgement, probably audible, e.g. a beep or some pre-recorded message or tune;
• [0029] (iii) user says the voice command, e.g. a name to be dialled;
• [0030] (iv) user engages in the call (using speaker phone operation)—or executes whatever other command has been pre-programmed.

[0031] Speaker phone operation with a mobile terminal requires a loud audio output and some form of echo control.